The Material Separation Process for Recycling End-of-life Li-ion Batteries
Liurui Li
Dissertation submitted to the faculty of the Virginia Polytechnic Institute and State University
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
In
Mechanical Engineering
Zheng Li
Michael W Ellis
Rui Qiao
Alex O Aning
9/25/2020
Blacksburg, VA
Key words: Battery Recycling, Material Separation, Automated Disassembly, Design of Experiment
Copyright (2020)
The Material Separation Process for Recycling End-of-life Li-ion Batteries
Liurui Li
Abstract
End-of-life lithium-ion batteries retired from portable electronics, electric vehicles
(EVs), and power grids need to be properly recycled to save rare earth metals and avoid any
hazardous threats to the environment. The recycling process of a Lithium-ion Battery
Cell/Module includes storage, transportation, deactivation, disassembly, and material recovery.
This study focused on the disassembly step and proposed a systematic method to recover
cathode active coating, which is considered the most valuable component of a LIB, from end-
of-life LIB pouch cells. A semi-destructive disassembly sequence is developed according to the
internal structure of the LIB cell. A fully automated disassembly line aiming at extracting
cathode electrodes is designed, modeled, prototyped, and demonstrated based on the
disassembly sequence. In order to further obtain the coating material, the extracted cathode
electrodes are treated with the organic solvent method and the relationship between process
parameters and cathode coating separation yield is numerically studied with the help of Design
of Experiment (DOE). Regression models are then fitted from the DOE result to predict the
cathode coating separation yield according to combinations of the process parameters. The
single cell material separation methodology developed in this study plays an important role in
the industrial application of the direct recycling method that may dominate the battery recycling
market due to its environmental friendly technology and high recovery rate regardless of
element type in the short future.
The Material Separation Process for Recycling End-of-life Li-ion Batteries
Liurui Li
General Audience Abstract
The bursting demand of lithium-ion batteries from portable electronics, electric vehicles,
and power grids in the past few years not only facilitate the booming of the lithium-ion battery
market, but also put forward serious global concerns: Where should these batteries go at their
end-of life and how should they be treated in a safe and harmless manner. As a metal enriched
“city mine”, end-of-life LIBs are expected to be properly stored, transported, deactivated,
disassembled, and recovered with sufficient safety precautions to prevent fire, explosion or any
hazardous emissions. This study focuses on the disassembly procedure and emphasized
automated battery disassembly techniques and the improving of material separation efficiency.
A disassembly sequence of the pouch cell is scheduled and optimized for the first time. To
realize the scheduled sequence, a fully automated pouch cell disassembly system is designed to
achieve semi-destructive disassembly of z-folded pouch cells. Fixtures, transporters and end-
effectors were prototyped and assembled into the modularized disassembly line which extracts
cathode electrodes as final product. Cathode electrodes as the most valuable component in a
LIB then need to go through multiple chemical-mechanical treatments to future separate
cathode coating and Al current collector. This study utilized DOEs to optimize the operating
parameters of the material separation process for Lithium cobalt oxide (LCO) coating and
Lithium iron phosphate (LFP) coating. Regression models are successfully established for yield
prediction with certain levels of control factors.
iv
Acknowledgement
Firstly, I would like to express my sincerest gratitude to my advisor, Dr. Zheng Li.
Without his professional guidance on both theoretical and experimental aspects of my research,
I could not have finished my Ph.D. study. His passion for research and creative research
intuition has inspired me profoundly. I feel extremely lucky to have Dr. Zheng Li as my
committee chair.
I also want to thank my committee members: Dr. Michael Ellis, Dr. Rui Qiao, and Dr.
Alex Aning for their guidance and encouragement on my research work.
I must thank Yingqi Lu, Tairan Yang, Panni Zheng, and Dayang Ge who I cooperated
with in the same research group during the past three years. They always spared no effort to
help me whenever I’m in trouble with my research or daily life. I could not thank more for their
selfless assistance.
Special thanks to my parents and my wife, Yuxian Ye. Having their company to share
happiness and sour is the best thing that could happen to me in my life.
v
Table of Content
Chapter 1 Introduction ............................................................................................... 1
1.1 Lithium-ion Battery Technology......................................................................................................... 1
1.1.1 Lithium-ion Battery Electrochemical Mechanism ....................................................................................... 1
1.1.2 Lithium-ion Battery Manufacturing ............................................................................................................. 4
1.1.3 Lithium-ion Battery Retirement ................................................................................................................... 7
1.2 State-of-the-Art Pretreatment Process in End-of-Life (EoL) LIB Recycling ................................... 10
1.2.1 Recycling strategies overview ............................................................................................................... 10
1.2.2 Module/Pack Disassembly ..................................................................................................................... 12
1.2.3 Material Separation .................................................................................................................................... 14
1.3 State-of-the-Art Material Recovery Process in EoL LIB Recycling........................................... 19
1.3.1 Pyrometallurgical Processes ...................................................................................................................... 19
1.3.2 Hydrometallurgical Processes .................................................................................................................... 20
1.3.3 Direct Recycling Processes ........................................................................................................................ 24
1.3.4 Comparison of Three Major LIB Recycling Methods ............................................................................... 26
1.5 Outline of this Dissertation ............................................................................................................... 28
Chapter 2 Disassembly Automation for Pouch Cells ..............................................31
2.1 Disassembly Sequence Planning ....................................................................................................... 33
2.1.1 Disassembly Mode Selection ..................................................................................................................... 33
2.1.2 Disassembly Precedence (DP) Confirmation ............................................................................................. 37
2.1.3 Modules Design ......................................................................................................................................... 41
2.2 Modules Design and Prototype ......................................................................................................... 43
vi
2.2.1 Pouch Trimming Module ........................................................................................................................... 43
2.2.2 Housing Removal Module ......................................................................................................................... 46
2.2.3 Electrode Sorting Module .......................................................................................................................... 48
2.3 System Integration and Testify ......................................................................................................... 51
2.3.1 Control Architecture .................................................................................................................................. 51
2.3.2 Concept Verification .................................................................................................................................. 53
2.4 Sorting Module Upgrade and Future Development .......................................................................... 56
2.4.1 Sorting Module Upgrade ............................................................................................................................ 56
2.4.2 Cyber-physical Closed-loop Controller and Preliminary Experiments ...................................................... 60
2.5 Conclusion ........................................................................................................................................ 66
Chapter 3 Cathode Coating Separation of Lithium Cobalt Oxide Battery ..............67
3.1 Experiment Setup .............................................................................................................................. 69
3.2 Placket-Burman Parameter Screening Experiment ........................................................................... 71
3.2.2 Placket-Burman Parameter Screening Experiment Design ........................................................................ 71
3.2.3 Parameter Screening Results ...................................................................................................................... 73
3.3 Taguchi DOE .................................................................................................................................... 76
3.3.1 Experiment Design ..................................................................................................................................... 76
3.3.2 Taguchi DOE Results ................................................................................................................................ 79
3.3.4 Linear Regression Model for Yield Prediction .......................................................................................... 82
3.4 Conclusion ........................................................................................................................................ 84
Chapter 4 Cathode Coating Separation of Lithium Iron Phosphate Battery .........85
4.1 Full-Factorial Experiment Setup ....................................................................................................... 85
vii
4.2 Result and Discussion ....................................................................................................................... 89
4.2.1 Full-factorial DOE Separation Yield ......................................................................................................... 89
4.3.2 Regression Analysis ................................................................................................................................... 92
4.3.3 Correlation Pre-confirm with Contour Plot and Latin Square .................................................................... 94
4.3 Conclusion ........................................................................................................................................ 97
Chapter 5 LFP Direct Recycling ..............................................................................99
5.1 EoL LFP Direct Regeneration........................................................................................................... 99
5.2 Coin Cell Assembly ........................................................................................................................ 104
5.3 Electrochemical Performance of the Regenerated LFP .................................................................. 106
5.4 Conclusion and Future Development .............................................................................................. 108
Chapter 6 Summary and Future Work ...................................................................109
6.1 Summary of Contributions .............................................................................................................. 109
6.2 Future Work .................................................................................................................................... 110
REFERENCES .......................................................................................................113
viii
List of Figures
Figure 1-1 Schematic illustration of a typical lithium cobalt oxide battery17. ................................ 2
Figure 1-2 Three categories of ESCs differenciated by cotiniuity of electrodes and separator33. .. 6
Figure 1-3 Solid-electrolyte interface development and aging on anode surface42. ....................... 9
Figure 1-4 General flow sheet of EoL LIB recycling processes (modified from 82) .................... 12
Figure 1-5 (a) Hybrid disassembly workstation and (b) Robot flange with electric screwdriver90.
....................................................................................................................................................... 14
Figure 1-6 Pretreatment processes summary for recycling EoL LIBs in industrial scale and lab
scale99. ........................................................................................................................................... 15
Figure 1-7 Process steps of a typical destructive material separation process106. ........................ 17
Figure 1-8 Experimental set-up for the ANVIIL process97. ......................................................... 18
Figure 1-9 Flow sheet for a typical hydrometallurgical process of LiCoO2 from spent LIB126. .. 21
Figure 1- 10 Material flow of the direct recycling process156. ...................................................... 24
Figure 1- 11 Direct regeneration process flow diagram of cathode material mixture81. .............. 25
Figure 1- 12 Comparison of different LIB recycling methods83................................................... 27
Figure 2-1 Configuration of the H605060 lithium-ion polymer rechargeable battery manufactured
by MTI Corporation. ..................................................................................................................... 35
Figure 2-2 Detailed 2D specification of the H605060 lithium-ion polymer rechargeable batteries
from MTI Corporation. ................................................................................................................. 36
Figure 2-3 Disassembly matrix of Z-folded pouch cells. ............................................................. 38
Figure 2- 4 Disassembly precedence graph and module function division. ................................. 41
Figure 2-5 Continuous process for recovery of cathode coating from end-of-life LIBs. ............. 42
Figure 2-6 (a) CAD design and (b) Prototype of the pouch removal module. ............................. 44
ix
Figure 2-7 Detailed design of the pinch roller conveyor set: (a) side view in CAD model, (b) top
view in CAD model, (c) assembly overview in CAD model, and (d) top view of prototyped pinch
rollers. ........................................................................................................................................... 45
Figure 2- 8 (a) Schematic and (b) Prototype of the pouch removal module. ............................... 46
Figure 2- 9 Design of the vacuum grip (a) front view and (b) side view .................................... 47
Figure 2-10 (a) Schematic and (b) Prototype of the electrode sorting module ............................. 49
Figure 2-11 Pneumatic securing unit for separator delivery. ........................................................ 50
Figure 2-12 Control architecture of the prototyped automatic recycling system ........................ 52
Figure 2- 13 Prototyped H605060 LIB disassembly system overview. ....................................... 53
Figure 2-14 Z-folded dummy cell assembly line following the size of a H605060 LIB: (a) Die cut
machine, (b) Al laminated film stamping machine, (c) ESC folding machine, (d) Heat sealing
machine, (e) Vacuum sealing machine, (f) Trimmed Al foils, (g) Stamped Al laminated film, (h)
Folded dummy ESC, (i) Dummy cell with one side edge and the top edge heat sealed, and (j)
Vacuum sealed dummy cell. ......................................................................................................... 54
Figure 2- 15 12 key frames from system testing record corresponds to (a) handling scenario 2, 3,
4, and 5 of the trimming module, (b) handling scenario 1, 4, 8, and 11 of the housing removal
module, and (c) handling scenario 1, 3, 4, and 5 of the electrode sorting module. ...................... 55
Figure 2-16 Vision-sensor network components: (a) FLIR S USB3 mono industrial cameras with
1.6MP resolution and 226 FPS, and (b) Tension sensor modified from strain gage based load
cell. ................................................................................................................................................ 57
Figure 2-17 Schematic of the vision-sensor network integrated electrode sorting module. ......... 58
x
Figure 2-18 System operating with EoL H605060 LIB: (a) Upgraded sorting system overview,
(b) Operating frame of industrial camera #2, (c) Operating frame of industrial camera #3, (d)
Operating frame of industrial camera #1, and (e) Reading from the load cell. ............................ 59
Figure 2-19 LabVIEW control unit Front User Interface: (a) Image acquisition section, (b) Load
cell reading section, and (c) Linear motion control section. ......................................................... 59
Figure 2-20 Control architecture of the cyber-physical closed-loop process monitor for the
electrode sorting module. .............................................................................................................. 60
Figure 2-21 Algorithm input (a) intact separator and (b) cracked separator. ............................... 61
Figure 2-22 Layered structure of CNN working principle183. ...................................................... 62
Figure 2-23 Examples of image classification result and probability value. ................................ 63
Figure 2-24 (a) Structure of the confusion matrix and (b) Confusion matrix of the preliminary
algorithm. ...................................................................................................................................... 65
Figure 3-1 Battery modules disassembled for DOE. .................................................................... 69
Figure 3-2 Equipment requirement (a) Battery module discharger, (b) High temperature oven,
and (c) Ultrasonic cleaner. ............................................................................................................ 70
Figure 3-3 Experiment flow with single cathode electrodes. ....................................................... 71
Figure 3-4 Main effects plot for fitted means of screening experiment yield. ............................. 76
Figure 3-5 Main effects plot for S/N ratio. ................................................................................... 81
Figure 4-1 (a)Tenergy 3.2V 2500mAh LFP (IFR26650P) power cell rechargeable battery and (b)
Typical internal structure of a cylindrical LIB. ............................................................................ 87
Figure 4-2 The appearance of (a) the original LFP cathode electrode and the appearance of the
LFP cathode electrode soaked in NMP under 90℃ after (b) 2h, (c) 3h, and (d) 4h. .................... 88
Figure 4-3 Main effect plot for separation yield. .......................................................................... 90
xi
Figure 4-4 Fully separated LFP cathode coating (a) naturally detached from the Al current
collector during soaking, and (b) coating expansion comparison with the Al current collector. . 91
Figure 4-5 The contour plot plotted from the result of (a) the full factorial experiments and (b)the
latin square sampling subset. ........................................................................................................ 97
Figure 5-1 Process flow of the EoL LFP regeneration process. ................................................. 101
Figure 5-2 The XRD patterns of as-purchased LFP from MSE Supplies, EOL cathode materials,
and recycled cathode materials sintered under 550℃ , 650℃, 750℃ within Ar&H2 atmosphere,
and recycled material sintered under 750℃ in Ar atmosphere. .................................................. 102
Figure 5-3 The SEM images of (a) as-purchased LFP from MSE Supplies, (b) EOL LFP,
regenerated LFP sintered under (c) 550℃ , (d) 650℃, (e)750℃ within Ar&H2 atmosphere, and
regenerated LFP sintered under (f)750℃ in Ar atmosphere. ..................................................... 103
Figure 5-4 Cathode electrodes preparation for coin cell assembly. ............................................ 105
Figure 5-5 (a) Coin cell assembly explosive view, (b) MSK-160E digital coin cell assembly
crimper, (c) assembled CR2032, and (d) CT2001A classic battery tester. ................................ 106
Figure 5-6 The charge/discharge voltage profiles at 0.1C (a) and rate performance (b) of cathode
active material mixtures sintered under different conditions. ..................................................... 107
xii
List of Tables
Table 3-1 Screening experiment parameters and levels. .............................................................. 73
Table 3-2 Plackett-burman screening experiment design and result. ........................................... 73
Table 3-3 Analysis of variance of screening experiment. ............................................................. 75
Table 3-4 Taguchi experiment parameters and levels. ................................................................. 78
Table 3-5 Pre-set process parameters. .......................................................................................... 78
Table 3-6 Taguchi L16 orthogonal array and the corresponding separation yield. ...................... 79
Table 3-7 Response table for S/N ratio. ........................................................................................ 81
Table 3-8 Taguchi experiment analysis of variance for S/N ratio. ............................................... 82
Table 3- 9 Summary table of the regression model for LCO separation yield. ............................ 83
Table 3-10 Linear regression model for predicting LCO separation yield verification test result.
....................................................................................................................................................... 84
Table 4-1 Full-factorial experiment parameters and levels. ......................................................... 87
Table 4-2 Separation yield of the cathode coating in 2 factors, 4 level full-factorial experiment.89
Table 4-3 Summary table of the regression model for LFP separation yield. .............................. 92
Table 4-4 ANOVA table for the regression analysis. ................................................................... 93
Table 4-5 Result of the non-linear regression model verification tests. ....................................... 94
Table 4-6 A LSD of the 2 factors and 4 levels experiment. ......................................................... 95
Table 4-7 Latin square design of 2 factors and 4 levels full-factorial experiment. ...................... 95
Table 4- 8 LFP separation yield of the selected Latin Square sampling subset. .......................... 96
1
Chapter 1 Introduction
1.1 Lithium-ion Battery Technology
1.1.1 Lithium-ion Battery Electrochemical Mechanism
Lithium was first introduced into batteries by M. Stanley Whittingham in 1970s 1 at
which time titanium sulfide and lithium metal were used as the electrodes. Unaffordable prices
of titanium sulfide and insufficient voltage prevented the commercialization of such
combinations. In 1980, John B. Goodenough demonstrated a lithium-ion secondary battery
consisted of Lithium cobalt oxide (LCO) as positive electrode and lithium metal as negative
electrode2. The innovated combination enabled stable LCO to act as lithium donor, which
allowed the negative electrode to be replace by lithium holders other than lithium metal. The
first generation of commercial lithium-ion battery that suits for industrial-scale mass production
was invented by Akira Yoshino in 19853. He introduced carbonaceous with a certain crystalline
structure as the negative electrode after polyacetylene proved to be instable and too low in
density. Till present, lithium-ion batteries have been dominating the market of portable
electronics and EVs and revolutionized our daily lifestyles. For this and so much more
conveniences the technology has brought us, the 2019 Nobel Prize in chemistry was rewarded
to John B. Goodenough, Stanley Whittingham, and Akira Yoshino for the development of
lithium-ion batteries 4.
A typical LIB consists of four essential components: positive electrode, negative
electrode, electrolyte, and separator (Figure 1-1). A positive electrode typically consists of a
metal oxide and conductive additives held onto an aluminum current collector by a binder.
Positive materials can be layer structured (LCO2), olivine structured (LiFePO4), and spinel
2
structured (LiMn2O4)5,6. Currently, negative electrodes on commercialized LIBs are mostly
made from graphite and other carbon material coated onto a copper current collector by the
same binder as the positive electrode. Ever since Rachid Yazami demonstrated the capability
of the graphite in reversibly intercalating lithium-ion within a LIB cell 7, graphite has been the
dominant material for negative electrodes because of its low voltage, modest volume expansion,
and high performance8,9. Electrolytes can be liquid, gel, or dry polymer10-12. As the most widely
adopted type, the liquid electrolyte is usually a lithium salt such as lithium hexafluorophosphate
(LiPF6 ), lithium hexafluoroarsenate monohydrate (LiAsF6), lithium perchlorate (LiClO4),
lithium tetrafluoroborate (LiBF4), and lithium triflate (LiCF3SO3) dissolves into organic
solvent blends of ethylene carbonate, dimethyl carbonate, or diethyl carbonate13. Additives such
as vinylene carbonate (VC) and Lithium bis(oxalato) borate (LIBOB) are selectively added to
help with the formation and stabilization of Solid-liquid Interface (SEI), improvement of
overcharge tolerance, or capability of Li cycling14. Separators are critical components in LIB
batteries that utilize liquid electrolyte. A separator is an ion-permeable porous membrane
consisting of conventional polyolefin coated by ceramic particles under certain circumstances
to improve thermal stability15,16.
Figure 1-1 Schematic illustration of a typical lithium cobalt oxide battery17.
3
The term “rocking-chair battery” was invented to describe the working principle of a
LIB: Lithium ions transporting back and forth by the electrolyte between cathode electrodes
and anode electrodes through separators18. As demonstrated in Figure 1-1, both electrodes allow
lithium ions to intercalate into and deintercalate out from their structures to form compounds
containing lithium atoms. During charging process represented by green arrows, positively
charged lithium ions deintercalte out from layered LCO (cathode) in an oxidation reaction when
energy is provided from external circuit (Eq.1-1 left to right). They diffuse towards graphite
(anode) in LiPF6 solution (electrolyte) and migrate through separators that insulate cathode and
anode electrodes. Simultaneously, electrons move from positive electrode to negative electrode
through the external circuit and enable lithium ions to intercalate into layered graphite (anode)
(Eq.1-2 left to right) to fulfill the energy storage process (Eq.1-3 left to right) during which
liquid electrolyte and external circuit performed as conductive media for lithium ions and
electrons accordingly. At this stage, LIBs are at high energy state and their open-circuit voltage
(OCV) highly depends on electrochemical potentials of both positive electrodes (anode) and
negative electrodes (cathode). When fully charged LIBs connected to external loads, the
discharging process executes backward reactions and the majority of lithium ions would be
transferred back to the positive electrode (cathode), thus lowering the potential and the energy
state of the cell. Noticing that anode and cathode actually change positions during charge and
discharge process. To avoid confusion for the rest of this dissertation, cathode will represent
metal oxides that act as actually cathode electrode during discharging process at the positive
terminal marked on the battery shell and anode will represent carbon based anode electrode
during discharging process at the negative terminal. The transition metal, cobalt, in Li1−xCoO2
at cathode oxidized from Co3+ to Co4+ during charging and reduced from Co4+ to Co3+ during
4
discharging. The overall charging reaction and discharging reaction shown in Eq.1-3 was used
in the first commercial LIB developed by Sony and Asahi Kasei in 1991.
The cathode half reaction (left to right: charging, right to left: discharging):
LiCoO2 ⇌ CoO2 + Li+ + e− Eq. 1-1
The anode half reaction (left to right: charging, right to left: discharging):
C6 + Li+ + e− ⇌ LiC6 Eq. 1-2
The overall reaction (left to right: charging, right to left: discharging):
C6 + LiCoO2 ⇌ LiC6 + CoO2 Eq. 1-3
The success of commercial LIB utilizing liquid electrolyte has everything to do with the
existence of a passive film namely solid-electrolyte interface (SEI) formed on anode surface.
Since anodes operate at voltages that are much lower than the electrochemical stability window
of electrolytes, a small amount of electrolytes tend to decompose and form this film during the
first few cycles of a freshly assembled LIB at the cost of some irreversibly consumed lithium
ions19. The formation of SEI not only prevents the further decomposition of electrolytes but
maintains sufficient permeability for lithium ions 20. Changes of SEI throughout the life cycle
of a LIB are considered as one of the major sources of LIB degradation at the anode21,22.
1.1.2 Lithium-ion Battery Manufacturing
The manufacturing process of LIBs can be divided into three stages: electrode
production, cell production, and cell conditioning. For LIB modules, an extra cell quality
sorting and module assembly process is needed23,24.
Most electrode production processes are based on wet coating and roll-to-roll
technique25,26. Active material, conductive additives, and polymer binder are firstly dry
(optional) or wet mixed depending on whether the polymer binder is pre-dissolved in the
5
organic solvent (N-Methyl-2-Pyrrolidone, NMP). After degassing and filtering, the
homogenous suspension is then continuously coated on both sides of the current collector
sequentially by the slot die. The aluminum (cathode) and copper (anode) current collector
should have high conductivity, tensile strength, and surface adhesion as well as low E-modulus
and thickness variation27. The coated current collector then goes through a dryer under a
moisture-free environment. The organic solvent (NMP) evaporated from the wet coating during
the drying process needs to be recovered and recycled for both cost and environmental concerns.
Finally, the dried electrodes are compressed between two rollers separated by a certain distance
to acquire desired electrode thickness. The compression step also enhances the energy density,
conductivity, and adhesion as well as decreases the porosity of the coated active material28,29.
The cell production stage firstly cuts the dried continuous electrodes and properly
insulates cathode and anode by polymer separator to form the electrode-separator compound
(ESC). Depending on the continuity of electrodes and separator, configuration of ESCs can be
divided into three categories as shown in Figure 1-2: stacking, winding, and Z-folding30. The
stacked ESC consists of both discrete electrode sheets and discontinuous separator, which is
suitable for relatively large format batteries. The uniformed mechanical load was applied
throughout the entire sheet so that thicker electrodes could be involved to improve energy
density. Nevertheless, discontinuous components raise the requirement for precision alignment
and obviously decreases productivity. The second ESC configuration is winding, which can
either be cylindrical or prismatic in shape31. Continuous electrodes and separator decrease the
assembly time to seconds without introducing any misalignment concerns. But the bending
nature of this ESC type and crack formation risk limit its energy density and electrode format,
especially electrode thickness32. Compared to two designs above, Z-folding ESC is a more
6
balanced configuration between productivity and yield. The continuous separator is folded into
zig-zag shape and discrete cathode and anode sheets are inserted in-between alternately. Since
electrodes insulated by the separator are stacked in a plane, the advantage of the stacking ESC
applies to Z-folding as well. The continuous separator also decreases the number of items for
assembly and the difficulty for alignment.
Figure 1-2 Three categories of ESCs differenciated by cotiniuity of electrodes and separator33.
Cathode and anode electrodes in assembled ESCs are then welded to metal tabs for
positive terminal and negative terminal correspondingly before being inserted to an open
housing23,31. The housing of LIBs can be divided into three main categories: cylindrical,
prismatic hard case, and prismatic pouch cell. Cylindrical cells have the highest productivity
but relatively low energy density when assembled into modules and packs due to insufficient
usage of the space. Prismatic hard cases provide the best protection for ESCs against external
mechanical impacts in trade of volumetric energy density and specific energy density. The
prismatic pouch cell utilizes formed aluminum laminated film as case to provide sufficient
7
insulation and protection to ESC. After filling in the electrolyte, housings will be closed and
sealed to finish the cell production stage.
The initial few cycles of charging and discharging of a newly assembled cell is called
the formation process34,35. SEI layers are expected to form due to decomposition of the
electrolyte at the surface of electrodes, especially anode. During this process, gaseous
electrolyte decomposition products such as ethylene (from ethylene carbonate), propylene
(from ethylene carbonate), or 𝐶𝑂2 gase 36,37 will accumulate inside sealed battery housings. Pre-
designed pressure relief valves enable cylindrical cells and prismatic hard-case cells to naturally
degas in the formation process while prismatic pouch cells rely on an additional pouch bag
attached on one side edge to absolve the gas and eventually cut off from the cell38. The aging
process is the final step of the cell production which is also the most time consuming step (2-3
weeks39). Cell aging performance will be used to identify short circuits and classify individual
cells according to their capacity, impedance, efficiency, etc. The classification result is of vital
importance to increase uniformity of battery modules, thus benefit equalization in battery
management system (BMS)40.
1.1.3 Lithium-ion Battery Retirement
Although unique advantages in power, energy density, self-discharge rate, and operating
temperature window 41 have helped LIBs widely adopted in EVs and energy storage plants in
recent years, their lifespan studies are still catching much attention in both academia and
industry. LIBs are expected to have a lifespan of 15 years for 42V battery systems and hybrid
electrical vehicles (HEV’s) and 10 years for electrical vehicles (EV’s) or up to 1000 cycles at
80% depth-of-discharge (DOD) according to United States Advanced Battery Council
(USABC)42. Apart from lithium intercalate/deintercalate reaction, massive side reactions such
8
as electrolyte decomposition, active substance dissolution, and lithium metal plating are
simultaneously proceeding within each charging-discharging cycle43-45. Thus the degradation
of a LIB is a complex process that may be caused by changes of anode, cathode, and electrolyte
throughout the entire life-cycle42,46,47.
At the anode side, the degradation mainly comes from changes at the SEI21,48-52. The
formation of SEI during the first few cycles of a newly assembled LIB decreases the total
amount of free Lithium and increases the impedance of the cell for the greater good of the anode
and electrolyte protection45,53. However as LIB cycles, electrolyte reduction reactions by the
anode continue at a low rate and gradually thicken the SEI thus causing further lithium loss and
impedance rise37. The growth of SEI also decreases the accessible surface area for lithium-ion
to pass through SEI and further increases the impedance54,55. Conditions such as high operation
temperature and high SOC accelerate such processes56,57. When LIB operates under low
temperature or high cycling rate, metallic Lithium tends to plate on the anode surface and result
in subsequent electrolyte decomposition by metallic Lithium. Consequently, capacity and
power of the LIB will fade due to loss of lithium and electrolyte58,59. As embedded Lithium
further accumulates, lithium dendrites start to form as shown in Figure 1-3. The growth of
lithium dendrites can pierce the porous separator and cause a short circuit that leads to thermal
runaway60,61. Meanwhile, the lithium intercalation and deintercalation processes are
accompanied with volume change of electrode material (around10% for graphite anode)62. High
cycling rates (fast charging for example) and state of charge induces extra mechanical strain
due to faster and larger volume change that may cause orientation of the graphite particle
compared to the original state and accelerate the fade63.
9
Figure 1-3 Solid-electrolyte interface development and aging on anode surface42.
Degradation from electrolyte mainly caused by electrolyte decomposition due to
abusing operation conditions of battery cells64. When overcharged to 4.5V or above (with
respect to Li/Li+), the electrolyte will be oxidized at the cathode and form insoluble substances
like Li2CO3 and LiF. For most commercial Li-ion, the degradation of the SEI layer will
accelerate at a cell temperature of 75 – 90°C65,66. The loss of SEI above such temperature will
lead to continuous reduction reaction between electrolyte and Li metal enriched anode.
Similar to anode, lithium intercalate/deintercalate also lead to volume change of the
cathode material, thus introducing mechanical stress and distortion of the crystal lattice42. Other
cathode changes may reduce the lifespan of a LIB include but not restricted to: cathode SEI
formation due to electrolyte oxidation67,68, Mn dissolution47,69, and high temperature
decomposition40,70.
10
1.2 State-of-the-Art Pretreatment Process in End-of-Life (EoL) LIB Recycling
As transition metals enriched city mines, EoL LIBs are of great value if treated properly.
For example, cobalt enriched LCO batteries have the potential recycling value of $8900 per ton
which is the highest among all common cathode materials while Lithium manganese oxide
(LiMn2O4 or LMO) batteries are lowest in recycling value at around $860 per ton 71. Currently,
solid waste landfills, waste-to-energy facilities, and recycling facilities are three major
destinations for EoL LIBs72. LIBs treated by solid waste landfill may bring severe
contaminations to groundwater through leachates, thus can pose a threat to the environment and
human health73. Waste-to-energy is a commonly found waste management method especially
in European countries. Though, energy can be generated out of combustion, hazardous gas
emissions and metal concentrated ash are inevitable drawbacks of dealing EoL LIBs within
waste-to-energy facilities. To maximize the value of each LIB components out of LIB waste
streams, recycling methods that are able to properly disassemble, separate, and reuse/reenergize
components of LIBs have drawn much attention in both industry and academia. This section
will give a general overview of the existing major recycling strategies followed by detailed
pretreatment process introduction.
1.2.1 Recycling strategies overview
For the convenience of interpretation, recycling procedures are divided into pre-
treatment processes (PTPs) and material recovery processes (MRPs) in this dissertation as this
study mainly serves the development of PTPs. Generally, there are three major types of MRPs:
pyrometallurgy, hydrometallurgy, and direct recycling72. Each MRPs raise unique requirements
to PTPs at different levels of complexity.
11
The primary goal of recycling LIBs is to separate and collect each components in high
purity so that they can be reintroduced into the production of new LIBs or other related products.
Typically, a LIB consists of around 25-30 mass% cathode, 15-30 mass% anode, 10-15 mass%
electrolyte, and 3-4 mass% separator 74,75, among which precious metals (Lithium (Li), Cobalt
(Co), and Nickel (Ni)) enriched cathode material attract the most attention for LIB recycling
pioneers. Thus the pyrolysis process that is aiming at burnout LIBs as a whole and only recover
valuable metals from the slag, namely pyrometallurgy process, has been widely adopted in
battery recycling companies such as Umicore, Accurec, Sony, Onto and Inmetco76. The
Pyrometallurgy processes merely require module discharging as the PTP which is the simplest
among the three major MRPs. The hydrometallurgy process mainly includes steps of selective
leaching, solvent extraction, and precipitation77,78. Thus to increase the efficiency of the MRPs,
destructive mechanical crushing and sieving PTPs are commonly found in hydrometallurgy
recycling plants. Recent years, as a wide variety of cathode materials have been commercialized,
hydrometallurgy processes can be restricted by lacking in adaptively to different battery types
and elements79. The direct recycling method re-energizes and reuses the cathode material
separated from electrodes with minimum changes to the active material’s crystal morphology80.
Cathode powders collected from PTPs go through simple solid-state synthesis with additional
LiOH, which also raises the complexity of PTPs since cathode electrodes are expected to be
separated from other components of LIBs with their integrity well preserved81. Figure 1-4 gives
a brief overview of three aforementioned strategies.
12
Figure 1-4 General flow sheet of EoL LIB recycling processes (modified from 82)
1.2.2 Module/Pack Disassembly
Before the EV transformation in ground transporting vehicles, the majority of end-of-
life LIBs came from portable electronics either in the form of individual cells or small packs.
As the EV market grows rapidly, sales of electric vehicles in 2017 alone may lead to 250000
tones and half a million cubic meters of module/pack waste LIBs83. Disassembling these
modules/pack manually can be hazardous to human operators and time-consuming due to
lacking in design for disassembly in the product design stage84. Thus automating robotic battery
module/pack disassembly systems aiming at eliminating safety risks for human workers and
reducing production cost has been studied in a number of research works 85-90.
Jan Schmitt 88 developed a robotic gripper system to rationalize the automated
disassembly processes of large scale lithium ion battery packs. The hardware structure, PLC
based control architecture, and the corresponding software were presented in detail. Other than
13
performing as a fixer and transporter of individual cells, the gripper system was also integrated
with functionalities to characterize a battery cell’s residual voltage and resistance which will be
further used in status evaluation in the disassembly process. The gripper system was later
introduced into a scenario-based disassembly system for automotive LIB systems as part of a
hybrid disassembly system that can benefit the effectiveness and viability of the disassembly
process85.
Kathrin Wegener 90 proposed an EV battery module/pack disassembly sequence for
recycling Audi Q5 hybrid system and introduced a KUKA robotic arm (Figure 1-5(b)) for
unscrewing in a hybrid human robot workstation (Figure 1-5 (a)). The locating function of the
robotic arm was investigated in two options: human demonstration and the camera-based auto
detection. The human training method has been widely adopted for the robot arms in the
industrial assembly tasks, which suits for situations with minimum uncertainties. As for the
disassembly of a battery module, deformation or corrosion from years of service may
compromise the human-robot collaboration. Previous research works 91-93 has demonstrated the
success of machine vision and machine learning in assisting high-precision and flexible robotic
assembly, thus Kathrin also proposed the automatic detection of bolts with machine vision
system to increase the flexibility and assure the accuracy of the bolt removing process.
Kathrin’s research reveals the major challenge of applying robotics and automation equipment
in EV battery module/pack recycling. The poorly structured operating environments caused by
lacking in design standard for battery module/pack as well as the unexpected ware in years of
service emphasized the importance of involving artificial intelligence (AI) for objects
identification and uncertainty handling. Though automated disassembly of the wasted EV
module/pack have been attempted in human-robot hybrid workstations, there is still a long way
14
to go before AI control algorithms and software take control of the battery module/pack
disassembly tasks.
Figure 1-5 (a) Hybrid disassembly workstation and (b) Robot flange with electric screwdriver90.
1.2.3 Material Separation
The material separation process aims at concentrating the valuable cathode components
for the subsequent process. The major challenge in this step is to overcome the tight adhesion
between cathode powders and Al current collector formed by PVDF binder94. The material
separation methods can be roughly categorized as destructive process or semi-destructive
process depending on the integrity of the essential components introduced in section 1.1.1 after
the separation. The destructive process mainly consists of steps of crushing, sieving, and
classifying, which has been widely adopted in most researches and industrial production
systems for further treatments of hydrometallurgical and pyrometallurgical processes 79,95-98.
The semi-destructive process usually involves manual/semi-automated separation of the
essential components (cases, electrodes, and separators), elimination of PVDF binder, and
15
subtraction of cathode active material. Such ‘delicate’ material separation strategy ensured the
purity of the collected cathode powder which mainly benefited the uprising direct recycling
strategy due to its process sensitivity to impurity metals 81,97.
Figure 1-6 Pretreatment processes summary for recycling EoL LIBs in industrial scale and lab
scale99.
Controlling the hazard potentials during the material separation process is of vital
importance to achieve truly ecological friendly recycling of retired batteries. The major hazard
potentials of LIBs within the recycling process can be divided into two categories:
fire/explosion hazard and chemical hazard. The contemporary applied carbonates of the
electrolyte, such as dimethyl carbonate (DMC), ethyl methyl carbonate (EMC), and diethyl
carbonate (DEC), are highly flammable and show flash points as low as room temperature
(between 16 and 33°C)100. Meanwhile, the residual energy tends to cause thermal runaway if
batteries were accidently shorted by the external circuit. The crushing step in the destructive
process leads to micro-short circuiting between anode and cathode fragments and also
contributes to temperature rising. Under this circumstance, LIBs can easily cause fires and
16
explosions with an additional oxidant 101. The chemical hazards are mainly toxic gaseous
substances. With the presence of aerial humidity, LiPF6 in the electrolyte decompose (Eq.1-4)
and release hydrogen fluoride (HF) which can immediately convert to highly corrosive and
toxic hydrofluoric acid upon contact with moisture.
𝑳𝒊𝑷𝑭𝟔 + 𝑯𝟐𝑶 → 𝑳𝒊𝑭 + 𝑷𝑶𝑭𝟑 + 𝟐𝑯𝑭 Eq. 1-4
Jan Diekmann 102 introduced a typical destructive material separation process as shown
in Figure 1-5. The conducted mechanical process is part of the LithoRec process which
combines electrical, mechanical, thermal, and, most importantly, hydrometallurgical treatments
aiming at recovering nearly all valuable materials from a retired battery system 103-105. The
process yield and separation efficiency was investigated by the process steps of first crushing,
first air-classification, second crushing, sieving, and second air-classification. The discharged
EV battery modules were first crushed in a rotary-shear machine with a 20 mm discharge screen.
Within the zig-zag-sifting air classifier, heavy parts, such as module shells, electric conductors,
and steel screws, were selectively separated from the mixed fragments according to the density.
The second crushing was introduced to further decrease the fragment size so that black mass,
which is a mixture of cathode active material, anode active material, and impurities, can be
separated from the current collector foils and separator by a vibration sieve with a mesh size of
500μm. Finally, a second air-classification extracted the current collector foil fragments from
the valueless separator. The presented work demonstrate an industrial scalable mechanical
separation strategy for retired battery modules or battery cells with a 75% material recycling
rate.
17
Figure 1-7 Process steps of a typical destructive material separation process106.
The destructive mechanical separation process mainly relied on brutal force and density
or magnetism difference of the components to assure the separation efficiency. Thanks to its
high automation potential, such processes have been widely combined with hydrometallurgical
processes. Meanwhile, the semi-destructive process that aims at breaking adhesion between
active powders and current collector foils by dissolving the binder in a solvent 107-109or
decomposing the binder 97 110 under high temperatures has also caught attention in the lab scale
research.
The binder dissolving method was first adopted by Contestabile M107 who treated the
crushed battery mix with NMP at 100℃ for 1 h to separate LiCoO2 from the support aluminum
current collector. Soon after, the method of applying ultrasonic energy to facilitate the solvent
separation of cathode active material and aluminum current collector was introduced by Jinhui
Li 111. Such mechanical (ultrasonic)-chemical (organic solvent) combined method was widely
adopted ever since 112-114. The more detailed relationship between process parameters and the
separation efficiency of the ultrasonic enhanced binder dissolving process was first studied by
18
Song X115 in 2017. Process parameters such as temperature, sonication time, solid-to-liquid
ratio, and solvent type were studied by controlling a single variable at a time. The result
indicated that temperature and sonication time have little impact on the separation efficiency of
the selected spend LiFePO4 LIBs and smaller solid-to-liquid ratio facilitates the coating
separation. Meanwhile, in order to reduce the cost and pollution, C M Toma116 introduced acetic
acid as the ultrasonic medium to replace organic solvent and the optimal value of molar
concentration of acetic acid for coating separation was studied by the single variable
controlling approach as well.
Figure 1-8 Experimental set-up for the ANVIIL process97.
The binder decomposing method takes the advantage of the lower decomposition
temperature of the PVDF compared to graphite, carbon black, aluminum foils, and copper foils.
Therefore, a heating step in the material separation process can be both convenient and effective
to eliminate the adhesion. Based on this fact, Christian Hanisch 97 proposed the adhesion
19
neutralization via incineration and impact liberation (ANVIIL) approach as shown in Figure 1-
8. A high temperature of 500℃ was adopted in the furnace chamber for PVDF decomposition.
The newly designed air-jet separator then separated the coating powder and the current collector
foils by lifting the electrodes towards the top lib with a high-speed air to create the impact stress
on foils and agglomerates. The separated coating powder can then pass through the fine sieve
with a mesh size of 50 μm and get collected for the follow-up procedures. The ANVILL
approach was evaluated using LiCo0.33Ni0.33Mn0.33O2 cathode electrodes and the results
compared to a destructive separation process indicate a higher yield and a significantly higher
purity.
1.3 State-of-the-Art Material Recovery Process in EoL LIB Recycling
1.3.1 Pyrometallurgical Processes
The pyrometallurgical process usually uses a high-temperature furnace to decompose
the metal oxides in a reducing atmosphere. There are three types of products generated from
the pyrometallurgical process: metallic alloy, slag, and gases. The alloy, consisting of Co, Ni,
Fe, and Cu, is the reduction product of the metal oxides from the cathode electrodes117,118. The
different elements in the alloy can be selectively extracted and reused through
hydrometallurgical or bioleaching in the follow-up processes. The slag typically contains
aluminum, manganese, and lithium119,120 , which can be used as construction materials such as
cement120. Gaseous products are produced from the electrolyte and binder decomposition at low
temperature (<150℃) as well as burn off of the polymers at high temperature.
Without the need for module disassembly, pretreatment process, and passivation
process, the furnace is able to handle individual LIBs from portable electronics or even LIB
modules from EVs, which makes a great contribution for simplifying operations and reducing
20
production costs on a large-scale. In addition, the overall safety risk of the pyrometallurgical
process is little since individual LIBs or LIB modules are treated in the extreme temperatures
surrounded by the reducing atmosphere. However, the drawbacks of the pyrometallurgical
process are equally obvious as its advantages. The energy consumption is extremely high due
to high temperature operation. Even though the burning of the electrolytes and plastics (40-50%
of the battery weight) compensate for energy consumption, extra instruments are needed to
control toxic gases such as HF and CO generated from this process. Organic electrolytes,
binders, lithium, and manganese are either burned off or trapped in the slag, thus unable to be
reused for battery manufacturing. Despite the aforementioned disadvantages, the
pyrometallurgical process still remains a frequently adopted approach in industries for its
operation simplicity, safety, and high-yield for transition metals such as Co and Ni76,82.
1.3.2 Hydrometallurgical Processes
The hydrometallurgical process consists of two major steps: leaching and extraction.
Unlike the pyrometallurgical process, hydrometallurgical processes could not directly treat
individual LIBs or LIB modules. Instead, the input of hydrometallurgical processes have to be
either the smelting slag produced by the pyrometallurgical process121 or the black mass
produced by material separation processes introduced in section 1.2.3. Compared to the
pyrometallurgical process, the hydrometallurgical process holds many advantages such as
higher metal recovery efficiency, lower energy consumption, minimal gas emission, and lower
capital cost122-124. Thus, hydrometallurgical processes are better at recycling EoL LIBs on an
industrial-scale in favor of both cost and environment.
Figure 1-9 shows a complete material flow of a hydrometallurgical process, which
mainly included steps of passivation, material separation, leaching, extraction, and
21
remanufacture. What’s worth mentioning in the material separation step is the application of
the alkaline leaching method, which can be considered as a special semi-destructive material
separation process. The manually extracted cathode electrodes were soaked in sodium
hydroxide (NaOH) to dissolve Al current collector foils (see Eq.1-5). After water washing and
filtration, cathode active could be effectively extracted 125. The high-temperature calcination
process was then needed to burn off the PVDF binders to further eliminate the adhesion between
cathode powders.
𝟐𝑵𝒂𝑶𝑯 + 𝟐𝑨𝒍 + 𝑯𝟐𝑶 → 𝟐𝑵𝒂𝑨𝒍𝑶𝟐 + 𝟑𝑯𝟐 Eq. 1- 5
Figure 1-9 Flow sheet for a typical hydrometallurgical process of LiCoO2 from spent LIB126.
The leaching step is the major step for recovering valuable metals in the
hydrometallurgical process. It converts the metals in spent cathode powders from solid state
into solutions for further purification and separation processes by using inorganic acid, organic
acid, alkali or bacteria solution. Several inorganic acids such as sulfuric acid (H2SO4)76,127-130,
hydrochloric acid (HCl)111,131-133, and nitric acid (HNO3)134-137 are commonly used as leaching
22
agents, among which HCl alone proved to have the best leaching efficiency. However, the
oxidation of HCl would produce 𝐶𝑙2 (see Eq. 1-6), which increased recycling cost due to the
need of both antisepticising equipment and emission treatment equipment in the recycling
plants. A reducing agent, such as hydrogen peroxide(H2O2)112,128,136,138, sodium hydrogen
sulfite (NaHSO3)139, or glucose (C6H12O6)140, was usually required for using H2SO4 or HNO3
as leaching agent. Among all combinations of possible leaching acids and reducing agents
studied111,141-143, the most common combination by far is H2SO4/H2O2144. The reducing agent
H2O2 is able to convert transition metals contained in the black mass to their divalent states in
order to gain better solubility in acid solutions. Thus the leaching of LiCoO2 by H2SO4/H2O2 is
able to convert the insoluble Co (III) into soluble Co (II) through reaction as shown in Eq. 1-7.
𝟐𝑳𝒊𝑪𝒐𝑶𝟐 + 𝟖𝑯𝑪𝒍 → 𝑪𝒍𝟐 + 𝟐𝑪𝒐𝑪𝒍𝟐 + 𝟐𝑳𝒊𝑪𝒍 + 𝟒𝑯𝟐𝑶 Eq. 1-6
𝟐𝑳𝒊𝑪𝒐𝑶𝟐 + 𝟑𝑯𝟐𝑺𝑶𝟒 + 𝑯𝟐𝑶𝟐 → 𝟐𝑪𝒐𝑺𝑶𝟒 + 𝑳𝒊𝟐𝑺𝑶𝟒 + 𝟒𝑯𝟐𝑶 + 𝑶𝟐 Eq. 1-7
Although inorganic acids proved to have very high efficiencies in cathode active
material leaching, especially Li (>99%) compared to pyrometallurgical processes, the
considerable secondary pollutions limited the industrial application of inorganic acids. On the
one hand, hazardous gas emissions such as Cl2 , SO3 , and NOxare common products in the
inorganic acid leaching process, which posed a threat to both environment and human operators.
On the other hand, the waste acid solutions from the inorganic leaching process must be
neutralized by NaOH or treated by other approaches before emission. Extra expenditure will be
needed for further disposal of waste emissions if inorganic acids were used as leaching agents.
Thus organic acid leaching has caught much attention in recent years for their greener practices.
Some organic acids such as citric acid (C6H8O7)145 and oxalic acid (C2H3O4)146, showed great
potential in LIB cathode material due to their recyclability, minimum secondary pollution, and
23
easy degradation properties. Weak acidity of organic acids and relatively slower reaction speed
restricted the industrial application of such approach.
The dissolved solution after the leaching process contained valuable metals include Li,
Co, Ni, Mn, Cu, and Al. The second essential step of the hydrometallurgical process is to
selectively extract these metals out of the solution in ways of solvent extraction, chemical
precipitation, or electrochemical deposition. Because of the complexity of the solution, two or
more methods were commonly combined to separate metals from the solution82,96.
Solvent extraction approach was widely adopted in hydrometallurgy processes and
proved to be highly effective for different metal ions. Extractants including diethylhexyl
phosphoric acid (DEHPA), di(2-ethylhexyl) phosphoric acid (D2EHPA), bis(2,4,4-trimethyl-
pentyl) phosphinic acid (Cyanex 272), 2-ethylhexyl phosphoric acid mono-2- ethylhexyl ester
(PC-88A), and diethylhexyl phosphoric acid (DEHPA) were used to separate metals from
leaching solutions120,122,147. Chemical precipitation approach was commonly combined with
solvent extraction approach to gain a higher separation efficiency. The existence of anions such
as C2O42−
from oxalic acid (H2C2O4 ), PO43−
from phosphoric acid (H3PO4 ), OH− from
NaOH or ammonium hydroxide (NH4OH), and CO32−
from sodium carbonate (Na2CO3) may
combine with the valuable metals in the leaching solution and form precipitates148-152.
Although destructive material separation processes were most commonly adopted
pretreatment strategy for hydrometallurgical processes, the black mass containing both anode
and cathode materials would complicate the follow-up processes. Thus, the semi-destructive
material separation processes that separate anodes and cathodes prior to material segregation
would greatly benefit the final yield and expenditure. Nevertheless, lacking in universal design
standards for individual LIBs and LIB modules made it extremely difficult to achieve
24
automated pre-sorting of anodes and cathodes in large-scale production. Thus neither
pyrometallurgical process nor hydrometallurgical process could provide a pure and simplified
material stream to fulfil the closed-loop battery manufacturing.
1.3.3 Direct Recycling Processes
Both pyrometallurgical and hydrometallurgical processes aimed at produceing pure
metal as their final product, thus the compounds of cathode materials were first broken down
to element level before being selectively extracted from the mixture. However, the lithium
deficient cathode active powders can be reincorporated into new cathode product with the
presence of Lithium sources (lithium hydroxide (LiOH) or lithium carbonate (LiCO3)) through
a simple thermal treatment81,115,153-155, thus minimum crystal morphology changes of metal
oxides are expected. This step was usually named regeneration or re-litigation while the process
adopted this step was categorized to direct recycling process. Within the step, lithium sources
would compensate the lithium loss of the cathode material caused by SEI layer formation,
material degradation, and insufficient discharging (over discharging may lead to copper
contamination on cathode electrodes).
Figure 1- 10 Material flow of the direct recycling process156.
25
A simplified material flow of the direct recycling process is shown in Figure 1- 10.
Cathode electrodes need to be extracted as a whole within the pretreatment process to avoid
cross contamination with other material, especially copper foil. The adhesion between Al
current collector foil and cathode coating could be eliminated by either dissolving Al in NaOH
solution or eliminating the PVDF binders. The recycled active powders such as LCO could then
get “repaired” by LiOH under high-temperature calcination as show in Eq. 1- 8 where x is equal
to the average percentage of lithium losses in cycled material. The direct recycling processes
could directly produce high-value new cathode powder out of recycled black mass, which
avoided complexed leaching and separation processes. Thus, this approach had great potential
in decreasing the production cost as well as total energy consumption for the LIB cathode
material production cycle157,158. Meanwhile, all components of LIBs (except separators) could
potentially be recycled and reused if treated by this approach (See Figure 1- 11), which further
increased the economic return in industrial application.
𝑳𝒊𝟏−𝒙𝑪𝒐𝑶𝟐 + 𝒙𝑳𝒊𝑶𝑯 → 𝑳𝒊𝑪𝒐𝑶𝟐 + 𝒙 𝟒⁄ 𝑶𝟐 + 𝒙 𝟐⁄ 𝑯𝟐 Eq. 1- 8
Figure 1- 11 Direct regeneration process flow diagram of cathode material mixture81.
26
Nevertheless, huge obstacles prevent the practical application of the direct recycling
processes in recycling plants, for the drawbacks of direct recycling process were as obvious as
its advantages mentioned above. Firstly, directly repairing cycled cathode material put forward
a much higher requirement to pretreatment processes, especially automated module
disassembly and cathode electrode extraction. Previous research mainly focused on industrial
application of the destructive material separation process94,159 and automation potential of LIB
module disassembly160-162. How to effectively achieve semi-destructive cathode electrode
extraction within an automated production environment remained a blank. Secondly, simple
solid-state calcination process suited cathode materials with simple composition (LCO and
LFP). For cathode materials consists of multiple metal oxides like NCM and NCA, sol-gel
method or co-precipitation were needed to resynthesize cathode materials 163. The lithium
deficiency would be compensated into a molecularly homogeneous level in solutions instead of
solid-phase from these two approaches. Thus, precision presorting of individual batteries
according to material type prior to pretreatment process was required especially for EoL LIBs
from portable electronics164,165. Thirdly, the success of regeneration step was extremely
sensitive to metal impurities in cycled cathode coating material. The existence of fine Al or Cu
particles would compromise the electrochemical performance of the regenerated cathode
powder. Thus, apart from resynthesized cathodes, other high-value-added products consisting
of transition metals utilized in LIBs have also been widely studied141,166-171.
1.3.4 Comparison of Three Major LIB Recycling Methods
Although the advantages and disadvantages of three major LIB recycling methods have
been mentioned above correspondingly, it is still beneficial to compare their technology
readiness in a more intuitive way as shown in Figure 1-12. Pyrometallurgical process is by far
27
the most mature recycling strategy for practical application. The capability of smelting large
amounts of individual LIBs and LIB modules without involving any pretreatment process for
material separation significantly decreases the process complexity and production cost.
However, obvious drawbacks such as high energy consumption, severe second waste
generation, low yield, and lacking in lithium recovery ability urged both recycling industries
and researchers to search for its replacement with equal simplicity and productivity.
Hydrometallurgical processes hold the intermediate position of all three approaches in almost
every aspect. It has shown great potential in decreasing secondary waste generation and
maximizing recycling rate of Co and Ni while keeping the process complexity at a reasonable
level. Thus the hydrometallurgical process will most likely dominate the recycling industries in
the foreseeable future. Meanwhile, the direct recycling process is prevailing in a wide range of
lab-scale researches and barely started industrial application. The highest material recovery rate
and lowest secondary waste generation makes it the seemingly the most profitable recycling
process. Nevertheless, lacking in the design standard of individual LIB and LIB modules and
sensitivity to metal contamination weakened its attraction to large-scale production where
automation could not be emphasized more.
Figure 1- 12 Comparison of different LIB recycling methods83.
28
As important energy storage devices, LIBs play a more important role in solving energy
and environmental issues than ever before. Consequently, the LIB production volume rapidly
increases year by year. With limited service time, massive amounts of LIBs need to be properly
treated at their EoL. However, rapid iteration of the cathode materials and lack of structure
design standards have posed huge pressure to the recycling end of the full product cycle.
Although there is little hope to find a certain recycling process that can treat all types of LIBs
with different elements and configurations simultaneously in a profitable way, substantial
efforts from research lab and related LIB manufactures are expected to treat as many LIBs as
possible at their end-of-life state not only for maximizing the commercial profit but also for
eliminating severe environmental hazards that EoL LIBs might lead to.
1.4 Outline of this Dissertation
This dissertation focuses on the semi-destructive pretreatment process development that
serves the direct recycling strategy. To make up the blank of material separation method for
large-scale direct recycling of LIB cells as introduced in section1.2, the concept of an automated
disassembly system for the individual LIB cells is designed, prototyped, and investigated for
the first time. The following cathode coating extraction process is developed and improved on
both lab-scale and intermediate-scale production with the help of a series of carefully designed
experiments. The central line of our effort is to develop a systematic approach to enhance the
adaptively of the direct recycling process toward the large-scale production. The goal of this
research work is to facilitate the industrial application of the direct recycling strategy since it
has the great potential to become the most eco-friendly and economically beneficial approach
by far as introduced in section 1.3. The detailed structure of this dissertation is as follows.
29
Chapter 2 introduces the design and prototype of an automated disassembly system
aiming at separating cell case, metal tab, cathode electrodes, anode electrodes, and separators
of a LIB pouch cell with minimum human intervention. The disassembly sequence plan
indicates that the entire disassembly line to be divided into three modules connected with
conveyors. The prototyped modular disassembly system was then testified by manually
assembled dummy pouch cells.
Chapter 3 takes a step forward and focuses on cathode coating material separation of
cathode electrodes utilizing organic solvent approach. High power LCO battery packs are
selected as research subjects. General process input operating parameters are first differentiated
by Plackett-Burman parameter screening experiments. The parameters proved to be significant
to final yield are then studied in detail by Taguchi experiments to generate a regression model
for yield prediction under different input parameter combinations.
Chapter 4 studies the cathode coating material separation process of cathode electrodes
from LFP cylindrical secondary batteries. The organic solvent soaking process is able to
completely separate the coating materials and the Al current collector foil without involving
any mechanical energy assistance. Theoretically, this innovative approach is able to achieve a
100% pure cathode coating material recollection. The process yield is studied by full-factorial
experiments in order to provide guidance for process parameters selection for further lab-scale
or even industrial scale production.
Chapter 5 initiates the regeneration study of EoL LFP cathode coating reclaimed from
Chapter 4. The property of the EoL LFP is first characterized by Inductively Coupled Plasma
Atomic Emission Spectrometer (ICP-AES), X-ray Diffraction Spectrometry (XRD) and
Scanning Electron Microscopy (SEM). The electrochemical performance of samples
30
regenerated under various sintering conditions are then conducted to identify the optimum
process parameters.
Chapter 6 summarizes the major contributions of this dissertation. Possible future
developments are also enclosed.
31
Chapter 2 Disassembly Automation for Pouch Cells
As governments have begun to set timetables for banning the production of internal
combustion engine vehicles globally, the estimated annual demand for lithium-ion batteries
(LIBs) from electric vehicles (EVs) in 2025 reaches 408GWh, while this number was merely
20GWh in 2016.172 Previous studies on cycling performance degradation of battery packs in
hybrid electric vehicles (HEVs) indicate a battery pack lifetime of only 4.5 to 14.5 years
depending on their operating conditions.173,174 The rapid growth of discarded LIB packs from
HEVs and EVs along with an increasing number of end-of-life LIBs generated by portable
electronics and energy storage plants will cause severe environmental and safety problems if
not treated properly.82 Meanwhile, the end-of-life LIBs is a potential resource of valuable
metals (e.g. Ni, Mn, Li or Co) while pressures have been imposed on the supply chain of these
materials already.175 Therefore, it is of vital importance to develop recycling methodologies
that are ecologically friendly and economically feasible for end-of-life LIBs in present time and
future.
The state-of-the-art of end-of-life LIB recycling methods mainly combines mechanical
pretreatments and metallurgy processes.144 Mechanical pretreatments comprises steps of
discharging battery packs, dismantling packs into cells, and separating materials of single cells.
Metallurgy processes that consist of pyro-, hydro-, bio-metallurgy are generally downstream
procedures of mechanical pretreatments.82 Despite the dominating role of metallurgical
processes in the industry, their hazardous gas emissions, acid waste, and high energy
consumption issues have always been barriers toward a truly sustainable closed-loop
recycling.176 In recent years, a direct regeneration approach that resynthesizes recycled cathode
powder with heat treatment proves to be feasible in lab scale research.81,115,177 Although such
32
approach claimes to be more ecologically friendly and energy conserving, it also brings up
much higher requirement to the material separation technique toward mass production.
Material separation step, the most critical process in mechanical pretreatments, mainly
refers to multi-level crushing, sorting, and sieving techniques in the industry because of their
high automation potential. The total coating material recovery rate following this destructive
crushing strategy is only 75% even in lab scale research. 102 However, if electrode sheets can
be separated and extracted with their integrity well preserved, the coating material recovery rate
reaches as high as 97.4% utilizing am ANVIL process.97 Meanwhile, the black mass yielded
from the destructive crushing strategy is a mixture of anode coating, cathode coating, and metal
impurities. For metallurgical processes that mostly break down the compound of cathode active
materials, metals can be recycled or discarded by the following melting or leaching processes.
Nevertheless, for direct regeneration processes, the existence of anode powders and mixed
metals in cathode powder increases the complexity of downstream recycling processes and may
even decreases the electrochemical performance of the final product.178
Hence, an automatic disassembly system is designed and prototyped specifically for
dismantling and separating cathode sheets, anode sheets, separators, and Al laminated film
housing from lithium-ion pouch cells in this paper. Compared to the destructive crushing
strategy that has been widely adopted in industry, this proposed system has a great potential to
achieve higher coating material recovery rate as well as yield purer cathode powder for
mechanical pretreatment processes in mass production. The material separation strategy and
mechanism design of this project are patented within 179.
33
2.1 Disassembly Sequence Planning
Disassembly sequence planning (DSP) is always the first step of any EoL product
recycling processes. Just as the solutions of a product assembly problem may not be exclusive,
one can disassemble a product with multiple choices of precedence on particular parts and
components. A better-designed DSP benefits the disassembly efficiency as well as the
expenditure, thus helps with making actual profits out of the recycling activities which is the
practical problem that most recycling researches and enterprises are struggling on. The DSP for
individual LIBs are divided into three steps: disassembly mode (DM) selection, disassembly
precedence (DP) confirmation, and disassembly objective planning. In the mode selection step,
the suitable disassembly mode needs to be selected based on detailed target structures and
requests of subsequent processes. Then, the disassembly precedence relationships between each
components are confirmed by the disassembly matrix and the disassembly precedence graph.
Finally, a rough modular allocation of disassembly objectives are determined prior to detailed
mechanism design.
2.1.1 Disassembly Mode Selection
Choosing a proper DM is the first step of starting the recycling of a product at its EoL.
Generally, there are two categories of DMs on which decisions have to be made:
complete/fractional disassembly and sequential/parallel disassembly. The complete
disassembly usually takes apart the assembly into the fundamental compositions, thus each
component can be properly reused or recycled. The fractional disassembly only targets the high-
value added parts of the assembly and the complexity of the total disassembly process as well
as the total cost can be minimized compared to the complete disassembly. As for the sequential
disassembly, components are disconnected with the main assembly one after another. On the
34
contrary, the parallel disassembly removes multiple parts within one step, which is usually
applied to assemblies with relatively more parts and easy-to-access connections.
Before selecting disassembly mode, designers need to have a thorough understanding
of the shape and internal structure of the target EoL LIB. The shape of lithium-ion secondary
cells can be divided into three main groups: cylindrical, pouch, and prismatic. Their differences
in housing material and ESC design lead to certain advantages and disadvantages in
manufacturing processes and applications. Relatively low production cost and highly simplified
packaging processes allow pouch cells being extensively deployed on EVs and portable
electronics.
The typical internal structure of pouch LIBs is an ESC sealed by stamped Al laminated
film (Figure 2-1). Anodes and cathodes are alternately stacked and electrically isolated by the
separator to form the ESC. Depending on the continuity of electrode sheets and the separator,
ESCs are classified into three styles: single sheet stacking, winding, and Z-folding. The single
sheet stacking style is commonly used in laboratory cell assembly while the winding style
proved to be the most productive design in industrial applications due to its continuous
electrodes and separators feeding strategy. In recent years, efforts to decrease the cycle time of
Z-folding process utilizing robots and customized mechanisms have been carried out,39,180
which result in its rapid technology transition to mass production scale. Meanwhile, Z-folded
ESCs with discrete electrodes and continuous separators require more complicated mechanisms
to disassemble than the other two types. Thus, this dissertation mainly focuses on the recycling
of end-of-life pouch LIBs with z-folded ESCs.
35
Figure 2-1 Configuration of the H605060 lithium-ion polymer rechargeable battery manufactured
by MTI Corporation.
Based on aforementioned requirements, H605060 lithium-ion polymer rechargeable
batteries manufactured by MTI Corporation are chosen as the study subject for the research
work shown in this chapter. These cells have a nominal capacity of 2000 mAh at 300th cycle
with 0.2C discharge rate. The C-rate measures the rate that a battery is discharged according to
its maximum capacity. A 0.2C rate means that the discharge current can discharge the fully-
charged battery in 5 hour. Thus, for the H605060 LIB with a capacity of 2000 mAh, 0.2C
requires a discharge current of 400 mA. The concept of the C-rate will be repeatedly mentioned
throughout this dissertation, particularly within Chapter 5. The outer dimensions of the
H605060 LIB are: 5.8 ± 0.5 mm (Thickness) , 50.0 ± 0.5 mm (Width) , and 60.0 ±
0.5 mm (Length) as shown in Figure 2-2. The customized fixtures, transporters, and end-
effectors are designed to accommodate the dimensions and the clearance of the H605060 LIB.
36
Figure 2-2 Detailed 2D specification of the H605060 lithium-ion polymer rechargeable batteries
from MTI Corporation.
From Figure 2-1, it is clear that both side edges of the Al laminated film are folded
tightly towards the center and the heat sealing of the Al laminated film is irreversible. Certain
part of the Al Laminated film need to be cut off from the main body to reveal the ESC sealed
inside of it. Meanwhile two metal tabs on the positive and negative terminal are welded together
with all cathode electrodes and anode electrodes correspondingly by resistance welding
machines or ultrasonic welding machines. It is both not practical to unweld the metal tabs and
not profitable to spend extra expenditure on equipment and processes to recycle two metal strips.
Thus the concept of the fractional disassembly is applied to the disassembly of the H605060
LIB. As for the second category of the disassembly mode, parallel disassembly is not necessary
for H605060 individual LIB because of the relatively simple internal structure and the small
part number. Such disassembly mode can be extremely helpful in LIB module disassembly
where large amounts of subassemblies and connectors are involved and the operating space is
37
sufficient for multiple end-effectors to work simultaneously. Thus the components of H605060
LIB are to be removed sequentially in the order of Al laminated case and the ESC.
2.1.2 Disassembly Precedence (DP) Confirmation
After determining the DM as partial disassembly and sequential disassembly, the DP
has to be confirmed by developing disassembly models. A proper DP helps the developer to
better understand the inter connection between parts and components and make decisions on
disassembly sequence accordingly. Two most commonly utilized tools in disassembly model
development are the disassembly matrix and the disassembly precedence graph, which will be
applied to H605060 LIB disassembly in this section.
A complete disassembly matrix describes the interference between each component in
the direction of the x, y, and z-axes within a 3D space independently, which is commonly
adopted in complete disassembly mode since all connections have to be properly removed so
that each component can be isolated. Here, we have chosen to adopt the partial disassembly
mode according to aforementioned reasons, thus a unified disassembly matrix is developed
regardless of the direction of the interference between components. Within a disassembly
matrix, if the component A in the 𝑖𝑡ℎ row does not have any connections with the component
B in the 𝑗𝑡ℎ column, the A can be removed from B freely. Meanwhile, B can also be removed
from A without any constraints. If this is the case, the location ( i , j ) in the disassembly matrix
will be marked as 0. On the contrary, if external forces are required to separate certain
components, then the corresponding location in the disassembly matrix will be marked as 1.
38
Figure 2-3 Disassembly matrix of Z-folded pouch cells.
To better illustrate the property of the connections, a 3-axes space is defined for the
H605060 LIB. Assume the width (50.0 ± 0.5 mm) of the H605060 LIB is the x-axis, the
length(60.0 ± 0.5 mm) is the y-axis, and the thickness (5.8 ± 0.5 mm) is the z-axis. The Al
current collector foil tabs on cathode electrodes are welded together onto the positive metal tab
along the z-axis to form a unified positive terminal for all cathode electrode sheets. Thus
location (1, 1) and (5, 1) in the disassembly matrix as shown in Figure 2-3 are marked as 1.
Similarly, Cu current collector foil tabs on anode electrodes are welded onto the negative metal
tab along the z-axis as well, which marks location (2, 2) and (6, 2) as 1. The separator is the
most complicated component to define and remove connections in H605060 LIB. It is Z-folded
by itself, thus restricting its own free movement along the z-axis. As cathode electrodes and
anode electrodes are alternatively inserted into the folded structure of the separator, their
mobility in the x-axis is restricted unless the separator is unfolded. Hence, location (3, 1), (3,
39
2), and (3, 3) are marked as 1. The Al laminated film is a critical component for successful
pouch cell sealing and packaging processes. It consists of Nylon (ON) outer layer, Al middle
layer, polypropylene (CPP) inner layer, and two binding layers in-between them. The fantastic
stamping formability allows the Al laminated film to perfectly accommodate different sizes of
the ESCs. The stamped Al laminated housing not only provides protection to ESCs but also
restricts the mobility of all components of ESCs in the x-y plane, thus location (4, 1), (4, 2),
and (4, 3) are marked as 1. Meanwhile, the CPP inner layer, acting as the heating sealing layer,
seals two side edges and the top edges of H605060 LIB when two contacting CPP inner layers
are pressed by the heating bar along z-axis, thus location (4, 4) is marked as 1. As for the
positive metal tab, it is welded with the Al current collector foil of every single cathode
electrode, thus location (5, 1) is marked as 1. The top edge of the H605060 LIB also seals
positive metal tabs with the assistance of the hot melt adhesive polymer tape that came with the
metal tabs. Thus, location (5, 4) is marked as 1. Same connections and restrictions apply to
negative metal tabs as well hence location (6, 2) and (6, 4) are also marked as 1. So far, all
existing position restrictions caused by connection have been identified and recorded in the
simplified DM as shown in Figure 2-3. The next step is to generate the DP graph to eliminate
or remove all connections marked in DM by properly allocate them in DP.
Partial disassembly mode enables us to bypass complex steps of the traditional DP
generation process and make parts or components removal decisions with more freedom
especially when the integrity of some components do not need to be preserved. Here, the
designed material flow is presented in the tree graph as shown in Figure 2- 4. The original EoL
H605060 LIB is placed at the very top of the precedence graph as a whole. Branches stretch out
from assemblies or sub-assemblies at higher level and point to the isolated parts/components or
40
subassemblies at the next level. Prior to the exposure of the ESC, the connections that seal the
ESC inside the stamped Al laminated film housing need to be removed. As discussed in section
2.2.1, it is neither practical nor profitable to open up three sealed edges and unweld the metal
tab. Thus the first step is to trim them off from the main body of the H605060 LIB and leave
the Al film housing covered ESC as the subassembly for the next level. The trimming line of
the side edges will be located between edges of the Al film housing covered ESC and the sealing
strip, thus connection (4, 4) in the disassembly matrix can be removed within this step. The
trimming line of the top edge will be located between the welding point on the metal tab and
the Al/Cu current collector foils that are covered by active material coating. That says the
uncoated tabs on the Al/Cu current collector foils will be trimmed as well as the unsealed part
of the double layered Al laminated films same as the side edges. Hence, the connections held
by the welding point between cathode electrodes and similar connections held between anode
electrodes are eliminated including the connection between Al laminated films formed by heat
sealing. Meanwhile, since the metal tabs are abandoned along with the front edge, all
connections related to metal tabs are eliminated during the front edge trimming. Thus the front
edge trimming will remove connection (1, 1), (2, 2), (5, 1), (5, 4), (6, 2), and (6, 4). The Al film
housing covered ESC will be further taken apart in the next step to completely free the ESC
from the constraint of Al laminated films. The housing will be opened and exposed ESC can be
extracted for key material separation step, which result in elimination of connections (4, 1), (4,
2), and (4, 3) in x-y plane. Finally the z-folded separator of the ESC subassembly will be
stretched into a continuous strip, so that cathode electrodes and anode electrodes attached to the
opposite sides of the separator can be collected by two sets of customized end-effectors. Thus
in this level, position restriction applied to cathode electrodes and anode electrodes from the
41
separator are eliminated. Along with the stretching of the separator, connection (4, 1), (4, 2),
and (4, 3) are removed from the disassembly matrix. At this point, the major functionality of
the automated disassembly system has been identified and defined for each precedence level
and all connections existing in the disassembly matrix have been properly allocated in the
disassembly precedence graph. Further separation of active material from the cathode
electrodes will be covered in Chapter 3 and Chapter 4.
Figure 2- 4 Disassembly precedence graph and module function division.
2.1.3 Modules Design
Figure 2-5 shows the continuous process for direct regeneration of cathode materials
recycled from end-of-life LIBs in our lab scale production line. As part of this direct
regeneration strategy, our proposed single battery disassembly system has a great potential to
42
ensure the automatic separation of Al laminated films, separators, cathode sheets, and anode
sheets with their integrity well preserved. Before being fed into the disassembly system, end-
of-life LIBs need to be fully discharged in salt water to avoid any explosions or fire hazards.
Then, three sealed edges along with metal tabs are to cut off from the core area sequentially.
The remaining folded Al laminated housing film needs to be stretched from both sides by
external forces in order to extract the ESC. Separators need to be unfolded and continuously
fed forward. For end-of-life LIBs, electrode sheets tend to attach on the separator due to the
surface tension of the electrolyte or the nature bound due to aging. Thus, specialized skiving
tools are needed to scrape cathode and anode sheets off from opposite sides of the separator.
Figure 2-5 Continuous process for recovery of cathode coating from end-of-life LIBs.
43
Modularization design is adopted to increase flexibility of the disassembly system. The
system is physically designed and built into three modules as shown in Figure 2-4 and Figure
2-5: pouch trimming module, housing removal module, and electrode sorting module. These
modules should achieve the automated disassembly of H605060 LIB in the precedence
designed in Figure 2-4 and finish three key steps encircled in Figure 2-5 correspondingly. All
mechanisms or toolsets involved in these three modules can be divided into three categories:
end-effectors, fixtures, and transporters. Each module contains several customized key
apparatuses to achieve the designed connection removing plan. These apparatuses are usually
called end-effectors. Special fixtures are also needed to hold the position of targeted assemblies
or subassemblies while end-effectors operate. Between each module, the remaining
subassemblies such as Al film housing covered ESC and ESC itself need to be transported
between fixtures located in adjacent modules. These apparatuses are defined as transporters.
The cooperating of end-effectors, fixtures, and transporters assure the success of connections
removing as well as material separation in the automated disassembly line. Detailed design of
these key apparatuses are introduced in the next section.
2.2 Modules Design and Prototype
2.2.1 Pouch Trimming Module
In this module, the front edge of the pouch that carries electrode tabs is cut off thus
separating each of the electrode layers from the current collecting structure. The opposing side
edge seals are also removed so that the Al laminated film housing can be fully separated from
the compound.
44
Figure 2-6 (a) CAD design and (b) Prototype of the pouch removal module.
As shown in Figure 2-6, the pouch trimming module consists of three main components:
the trimming blade set, the trimming base set, and the conveyor roller set. The trimming blade
set is a triple-way aluminum frame with heavy duty breakaway blades fixed in their grooves
correspondingly. Each blade is 25-degree tilted from the horizontal direction to decrease the
cutting resistance force as well as maintain the cutting speed. The lower trimming base
cooperates with the trimming toolset in the trimming process by providing solid support to
double-layered Al laminated film. The clearance between trimming blades and edges of the
lower trimming base is controlled under 0.3mm while edges of the upper trimming base keep a
0.5mm distance from trimming blades to protect ESCs and avoid unnecessary cutting resistance
force. Two conveyor roller sets, each consisting of five powertrain gears and three 25A
polyurethane rollers as shown in Figure 2-7, are assembled inside both lower and upper
trimming base along with rotary shafts and ball bearings. These rollers with rough surface
45
texture can hold the pouch cell in position as well as transport the pouch without slippery. Thus
within this module, the triple-way trimming blade set and the trimming base set act as end-
effectors and the pinch roller sets take the responsibility of both fixture for the trimming process
and transporter to deliver the pouch cell into the trimming position as well as move the trimmed
subassembly towards the next module.
Figure 2-7 Detailed design of the pinch roller conveyor set: (a) side view in CAD model, (b) top
view in CAD model, (c) assembly overview in CAD model, and (d) top view of prototyped pinch
rollers.
The handling scenario of the pouch trimming process is as follows:
1. Feed fully discharged H605060 pouch cell into the trimming base.
2. Conveyor roller sets transport the pouch into the trimming position.
46
3. The trimming toolset moves down towards the trimming base and cut off three edges
of the housing.
4. Linear motion stage transport the trimming base forward.
5. Conveyor rollers deliver the trimmed pouch towards the next module.
2.2.2 Housing Removal Module
In this module, the remaining Al laminated film housing is to get peeled off from the
trimmed pouch. Hence, the recovery of Al laminated film from the end-of-life LIBs is
accomplished in this module. Figure 2-8 (a) demonstrates key procedures to stretch the housing
and extract the ESC while Figure 2-8 (b) shows the prototype of this module.
Figure 2- 8 (a) Schematic and (b) Prototype of the pouch removal module.
47
The housing removal module consists of three gripping apparatuses with specialized
functionalities: the transporting grip, the vacuum grip, and the clamping grip. The transporting
grip installed on a rotary base is designed to hold and transport the trimmed pouch and the ESC.
Distance between two flat arms at their parallel position equals the thickness of the targeted
pouch core. As shown in Figure 2-9, the vacuum grip is equipped with a height-adjusting
vacuum cup on each arm driven by 12V DC air pumps. These bellows suction cups are made
of translucent silicone, which provide sufficient adaptability to the uneven surface of Al
laminated film housing. The clamping grip is a fixture with a stationary flat base and vertically
movable top clamp. The width and the depth of the groove on the movable top clamp are
compatible with the size of targeted pouch cells in order to support and guide the ESC while
the vacuum grip peeling off the trimmed housing.
Figure 2- 9 Design of the vacuum grip (a) front view and (b) side view
48
The handling scenario of the pouch removal process is as follows:
1. The transporting grip extracts the trimmed pouch out of the trimming base.
2. Rotary base spin 90 degrees to feed the trimmed pouch into the clamping grip.
3. The clamping grip clamps the untrimmed side of the pouch core.
4. Rotary base spin 45 degrees to move the clamping grip away from the operation path
of the vacuum grip.
5. The vacuum grip closes the jaw and vacuum cups secure the upper and lower side of
the Al laminated film housing.
6. Jaws of the vacuum grip peel back upper and lower side of the Al laminated film
housing until the front edge of ESC fully exposed.
7. Rotary base spin back 45 degrees.
8. The transporting grip grasps the ESC.
9. Jaws of the vacuum grip further open the Al laminated film housing to its 180-degree
configuration while the untrimmed side of the housing pushes the electrode-separator
compound into the transporting grip.
10. The Al laminated film housing fell to the waste-recycling stream after the 12V DC
pump shut down.
11. The transporting grip delivers the ESC into the next module.
2.2.3 Electrode Sorting Module
The separation of cathodes, anodes, and separators is a critical process for any lithium-
ion LIBs recycling processes. It directly influences the purity and recovery rate of the black
mass. Our proposed electrode sorting strategy extracts cathode sheets and anode sheets
respectively without applying destructive forces. By automatically stretching and feeding the
49
Z-folded separator, cathode sheets and anode sheets attached on opposite sides of the separator
are scraped off by specialized tool sets as the schematic shown in Figure 2-10 (a). Since
commonly used PVDF binder can be removed by either dissolving in organic solvents or
decomposing at temperatures above 400℃ 102, multiple combinations of chemical, thermal and
mechanical treatments are then available for breaking the adhesion between Al foil and cathode
coating. The prototype of this module is shown in Figure 2-10 (b).
Figure 2-10 (a) Schematic and (b) Prototype of the electrode sorting module
50
The electrode sorting module consists of four apparatus: the vacuum conveyor, the
guiding posts, the skiving blades, and the pinch roller set. The vacuum conveyor is a height-
adjusting vacuum cup integrated on the X-Y motion platform. The 12V DC air pump enables
the vacuum cup to carry the top layer of the separator longitudinally to pinch rollers. Similar
pneumatic securing units as shown in Figure 2-11 are also applied to the vacuum grip
prototyped for the housing removal module. These urethane pinch rollers with 35A durometer
are able to feed the separator forward continuously while stainless steel guiding posts and break-
away skiving blades work as pairs to scrape electrodes off from the separator. After the 1st
guiding post pressing down on the soft separator, the front edge of the electrode sheets will
detach from the separator because of cathode sheets’ lower compliance to deformation. As the
separator is still rolling forward, the 1st skiving blade can seek its way through the gap between
the front edge of cathode sheets and the separator. Cathode sheets will then fall into the
collecting bin due to gravity. Anode sheets attached on the upper surface of the separator are
scraped off by the 2nd guiding post and skiving blade in a similar sequence.
Figure 2-11 Pneumatic securing unit for separator delivery.
51
The handling scenario of the pouch removal process is as follows:
1. The vacuum conveyor secures the top layer of the Z-folded separator and delivers it
through the pinch roller set.
2. The pinch roller set squeeze the top layer of the separator as vacuum conveyor release
vacuum.
3. The 1st guiding post and the 2nd skiving blade drops vertically to press down the
separator.
4. The pinch roller set continuously feeds the separator forward until the separator film is
fully paid off.
5. Components of the electrode-separator compound got separated into three collecting
bins correspondingly for further treatments.
2.3 System Integration and Testify
2.3.1 Control Architecture
A total of 11 stepper motors, 22 limit switches, 3 servo motors, and 3 pumps are installed
in the prototyped disassembly system. The realization of the designed handling scenario for all
three modules highly depends on an integrated control architecture. Therefore, a LabVIEW
platform is chosen to automate stepper motors, limit switches, servo motors, and vacuum
systems. This system-design platform offers great flexibility for design modification, which
will benefit the future optimization of this project. The derived control architecture of the
prototyped automatic disassembly system is as shown in Figure 2-12.
52
Figure 2-12 Control architecture of the prototyped automatic recycling system
Stepoko motion control boards (MCD) with GRBL firmware are used as local
controllers for stepper motors and limit switches. Each module is equipped with one to two
MCDs to achieve designed linear motion as well as relocate toolsets to their initial position.
Meanwhile, all 3 server motors and 3 DC vacuum pumps are controlled by one Arduino Mega
2560 R3 board. The LabVIEW program which served as a system controller for the ensemble
shown in Figure 2-13 is able to supervise the hand-off functions between modules and to
indicate event timing for serial operation of end-effectors, fixtures, and transporters.
53
Figure 2- 13 Prototyped H605060 LIB disassembly system overview.
2.3.2 Concept Verification
To assure the safety of the operators at the current prototyping stage, dummy pouch
cells are treated to verify the feasibility of the designed disassembly system. In dummy pouch
cells, non-toxic materials replaces key components of functional LIBs with similar physical
properties. The outer dimension of the dummy pouch cells is 50.0mm*60.0mm*5.8mm with
the tolerance of +/- 0.5mm. These values followed the size of H605060 LIB which will be tested
for the future development of this system.
54
Figure 2-14 Z-folded dummy cell assembly line following the size of a H605060 LIB: (a) Die cut
machine, (b) Al laminated film stamping machine, (c) ESC folding machine, (d) Heat sealing
machine, (e) Vacuum sealing machine, (f) Trimmed Al foils, (g) Stamped Al laminated film, (h)
Folded dummy ESC, (i) Dummy cell with one side edge and the top edge heat sealed, and (j)
Vacuum sealed dummy cell.
A pneumatic semi-automated Z-folded LIBs assembly line utilized to assemble dummy
cells is as shown in Figure 2-14. 300 micron thick paper cardboards are first trimmed by the die
cut machine. Along with an Al/Cu foil tab attached to the right position, these cardboards act
as the replacement of double-coated current collectors. For the Al laminated film stamping
machine, the depth of the stamping mode is adjusted to 5.6mm so that the overall thickness of
the dummy cell can be assured with the 0.1mm thick Al laminated film. An example the Al
laminated film stamping process is as shown in Figure 2-14 (g).The ESC folding machine
allows us to precisely assemble the dummy ESC by alternatively stacking cathode dummy
electrodes and anode dummy electrodes between the Z-folded separator. The dummy ESC will
then be placed into the stamped Al laminated film housing and have its side edge and top edge
55
sealed by the regular heat sealing machine as shown in Figure 2-14 (i). Diluted 3M liquid glue
is then filled inside the dummy cells from the open side edge to mimic the adhesion between
electrodes and the separator. The metal tab welding process is skipped for the dummy cell
manufacturing process since the trimming line at the front edge is located on top of Al current
collector foils and the welding points will be abandoned along with the trimmed front edge.
Finally, the dummy cell is placed inside the vacuum chamber of the vacuum sealing machine
(Figure 2-14 (e)) to drain the air out of the housing and heat seal the remaining side edge.
Figure 2- 15 12 key frames from system testing record corresponds to (a) handling scenario 2,
3, 4, and 5 of the trimming module, (b) handling scenario 1, 4, 8, and 11 of the housing removal
module, and (c) handling scenario 1, 3, 4, and 5 of the electrode sorting module.
Sealed dummy cells are then disassembled by the prototyped automatic disassembly
system. Figure 2-15 shows four key frames from the testing records of each module. Handling
scenario 2, 3, 4, and 5 of the trimming module are sequentially shown in Figure 2-15 (a). The
56
material flow of the single frame in Figure 2-15(b) and Figure 2-15(c) is from the right-hand
side to the left-hand side for clearer interpretation. Four frames in Figure 2-15 (b) correspond
to handling scenario 1, 4, 8, and 11 of the housing removal module. Recycling of the Al
laminated film is accomplished in the fourth frame of Figure 2-15 (b). The Al film housing
would fall off from the bellows cup after the DC pump release vacuum. Four frames in Figure
2-15 (c) correspond to handling scenario 1, 3, 4, and 5 of the electrode sorting module. As
shown in the fourth frame of Figure 2-15 (c), it is clear that the blue cardboards representing
cathode sheets, the green cardboards representing anode sheets, and the separator have been
successfully separated and stored in three locations.
The success in testing dummy cells manufactured following the assembly process of a
functionally H605060 LIB indicates the design rationality of the disassembly precedence and
customized apparatus. To be able to automatically disassemble the EoL H605060 LIB with the
prototyped disassembly system, a glove box with the ability to treat hazardous gas emission is
needed to keep the entire recycling process under inert atmosphere. This part of the work will
be covered in the project future development restricted by glove box hardware availability.
Instead, I take a step further in upgrading the prototyped system with industrial vision cameras
and sensors to monitor the disassembly process, thus allowing the control system to be more
intelligent in detecting system failures and act correspondingly. Some preliminary works on
sorting module upgrades are covered in the next section.
2.4 Sorting Module Upgrade and Future Development
2.4.1 Sorting Module Upgrade
The sorting module is rebuilt and upgraded within a fume hood so that the electrode
sorting process of a real ESC from H605060 LIB could be recorded and analyzed accordingly.
57
Apart from the original mechanism assembled in the prototyped disassembly line, three
Blackfly S USB 3 mono industrial cameras with 1.6MP resolution and 226 FPS as shown in
Figure 2-16 (a) are integrated. These cameras integrate the industry’s most advanced sensors
within a 29mm*29mm*30mm ice-cube. Both automatic and precise manual control modes over
image capture and on-camera pre-processing are available, thus enabling us to record videos
and capture images with trigger signals from LABVIEW. A load cell, which is a transducer that
measures force and outputs the force signal as an electrical signal, is also integrated into the
sorting module. The selected load cell as shown in Figure 2-16 (b) uses strain gauge to detect
load changes and hydraulic or pneumatic load cells are also favorable choices. The strain gage
type load cell usually integrates four strain gauges in a Wheatstone bridge. Such bridge circuits
originally have two balanced legs. When external load deforms the strain gauge, the electrical
signal changes can be captured by HX711 24bit precision ADC module that is connected to
LABVIEW via Arduino Mega 2560 R3 board.
Figure 2-16 Vision-sensor network components: (a) FLIR S USB3 mono industrial cameras with
1.6MP resolution and 226 FPS, and (b) Tension sensor modified from strain gage based load
cell.
58
Figure 2-17 Schematic of the vision-sensor network integrated electrode sorting module.
Figure 2-17 shows the schematic of the vision-sensor network integrated electrode
sorting module with minimum changes on the separator feeding mechanism. The original pinch
roller toolset is replaced by an initiative roller for the convenience of recording the load change
on the roller which equals the resistance force applied to the separator from guiding posts and
skiving blades. The overview of the upgraded sorting module is as shown in Figure 2-17. The
first industrial camera focuses on the ESC unfolding area as indicated in Figure 2-18 (d). The
second industrial camera monitors the first guiding post and the first skiving blade from the
above so that the separation process of the electrodes positioned on top of the separator can be
monitored as shown in Figure 2-18 (b). Ideally, the third industrial camera should focus on the
separation process of the electrodes attached to the bottom of the separator. However due to the
space restriction, the camera is set to 45° tilted from the horizontal direction and from Figure
2-18 (c). Any electrodes failed to be skived off from the separator can be identified through the
semi-transparent wet separator. The output of the load cell is as shown in Figure 2-18 (e), the
negative value between 5s and10s at the very beginning indicates a calibration step by applying
a standard 100g weight to the load cell.
59
Figure 2-18 System operating with EoL H605060 LIB: (a) Upgraded sorting system overview,
(b) Operating frame of industrial camera #2, (c) Operating frame of industrial camera #3, (d)
Operating frame of industrial camera #1, and (e) Reading from the load cell.
Figure 2-19 LabVIEW control unit Front User Interface: (a) Image acquisition section, (b) Load
cell reading section, and (c) Linear motion control section.
60
The control UI constructed within LabVIEW is as shown in Figure 2-19. The image
acquisition section as encircled in Figure 2-19 (a) allows me to integrate as many industrial
cameras as needed into the control software, so that future upgrades on the entire prototyped
disassembly system can be convenient. Figure 2-19 (b) displays the real time reading from the
load cell with the auto scale adjusting function. Figure 2-19 (c) is the linear motion control
section integrated with functions of the position restoration, precise linear positioning, and
arbitrary speed adjustment. The upgrade of the vision-sensor network for the electrode sorting
module initiates the study of the cyber-physical closed-loop LIB disassembly system. The
cyber-physical concept and some related preliminary works are introduced in the next section.
Figure 2-20 Control architecture of the cyber-physical closed-loop process monitor for the
electrode sorting module.
2.4.2 Cyber-physical Closed-loop Controller and Preliminary Experiments
The major purpose to design and implement a cyber-physical closed–loop controller in
EoL LIB disassembly system is to deal with the uncertain condition of the EoL LIBs. Even
LIBs with the exact same dimensions and electrochemical parameters, their conditions at EoL
61
can be very different due to their unique cycling conditions. Thus, a closed-loop supervision &
online machine parameter adjustment system will be needed. Figure 2-20 gives the overall
architecture of the cyber-physical closed-loop controller that consists of the physical part and
the cyber part. The physical part can be directly transferred from the prototyped disassembly
system introduced in the previous sections with an integrated vision-sensor network. The cyber
part which will mainly be established within Matlab or Python environments include two steps:
defects diagnosis and defects elimination actions. Take the electrode sorting module as an
example, the separator from an EoL LIB can be fragile and minimum cracks as shown in Figure
2-21 will lead to the breakage of the separator and eventually compromise the sorting process.
If such cracks can be recorded by the industrial cameras once they appear, machine learning
algorithms will be able to identify them by classifying the captured images. With the proper
physical model or complete experiment design, corresponding parameters can be adjusted to
either decrease the pulling force applied to the separator or simply shut down the operation and
call for human intervention. Thus, the implementation of this cyber-physical closed-loop
monitor will enable the machine to cognize the possible uncertainty and apply necessary
parameter adjustment correspondingly in real time. An example of detecting defects as shown
in Figure 2-21 and adjusting machine parameters according to the physical model of the
apparatus is demonstrated in the rest of this section.
Figure 2-21 Algorithm input (a) intact separator and (b) cracked separator.
62
Convolutional neural network (CNN) is used to identify the defects by continuously
classifying images captured with industrial cameras integrated during the electrode sorting
process for it provides promising results in image classification tasks by automatically
discovering the interconnections existed between disordered input181. Thus CNN, as a type of
deep neural networks, is quite often utilized in computer vision related analysis182. Figure 2-22
demonstrated the layered structure of CNN working principle. Input images, held as raw pixel
values, first go through the feature learning process where features of the images are identified.
The feature learning process usually consists of three types of layers: convolution layers, RELu
layers, and pooling layers. These layers will compute the output of neurons connected to local
regions from the original input images, apply elementwise activation functions, and down
sample the volume along the spatial dimensions. The classification process will further down
sample the volume of the output within the flatten layer. The fully connected layer is applied to
connect all features learned, hence all neurons are now connected. A softmax function is then
involved to calculate the probability of each label. The label with maximum probability value
will be the final classified category of the input image. In this way, CNN successfully transform
the input image layer by layer from pixel values to the final classified category.
Figure 2-22 Layered structure of CNN working principle183.
63
Two image labels for the cracking defect are defined as “new” and “crack” in the
preliminary CNN classification algorithm development. Around 750 images for each label are
used as the training data set and 350 images for each label are used as the validation set. Images
are downsized from the resolution of 4032*3024 to 200*150 to minimize the model training
time. Hyperparameters in a CNN model, such as learning speed, depth of the neuron network,
and batch size, determines the neuron network structure and setting. These parameters need to
be tuned to the best possible condition based on the classification accuracy of the validation set
before initiating the model training utilizing the training set. Here the Bayesian hyperparameter
tuning method is adopted. As shown in Figure 2-23, the trained CNN model will label each
image with the highest probability value as the classification result.
Figure 2-23 Examples of image classification result and probability value.
64
The trained CNN model is then testified by an 1120 mixed image pool containing
images from both categories. A confusion matrix is utilized to demonstrate the effectiveness of
trained CNN models. The matrix is a summary of prediction results on the classification
problem by giving us an insight of the errors being made by the CNN model. Figure 2-24 (a)
gives the basic structure of a 2*2 confusion matrix which is also the matrix size for cracking
defect detection since only two labels are involved. P in predicted class and actual class
represent that the observation is positive and N represent the negative observation. Here in the
cracking defect detection task, P represents the “crack” separator and N represents the “new”
separator. True Positive (TP) represents a positive observation and a positive prediction. True
Negative (TN) represents a negative observation and a negative prediction. Both TP and TN
are favorable results for the observation that agrees with the prediction. False Negative (FN)
represents a positive observation but predicted as negative and False Positive (FP) represent a
negative observation but predicted as a positive. Thus, FN and FP are major sources of
inaccuracy and should be avoided as much as possible. With TP, TN, FN, and FP available, the
accuracy of the CNN classification model can be calculated by Eq.2-1. With the confusion
matrix of the preliminary algorithm as indicated in Figure 2-24 (b), the accuracy of the trained
CNN model proved to be 98.12%, which is a satisfying result.
𝑨𝒄𝒄𝒖𝒓𝒂𝒄𝒚 = (𝑻𝑷 + 𝑻𝑵) (𝑻𝑷 + 𝑻𝑵 + 𝑭𝑵 + 𝑭𝑷)⁄ = (𝑻𝑷 + 𝑻𝑵) (𝑷 + 𝑵)⁄ Eq.2-1
65
Figure 2-24 (a) Structure of the confusion matrix and (b) Confusion matrix of the preliminary
algorithm.
The last step to fulfill the cyber-physical closed control loop is to enable the real time
machine parameter adjustment based on the physical model of the mechanism so that detected
defects can be contained or even eliminated. After detecting the “crack” defect, a series actions
can be taken to decrease the tension applied to the separator along its feeding direction. These
actions may include decreasing the separator tilting angle, decreasing feeding speed of the
separator, and increasing the blade tilting angle, which need to be further verified by the
physical model of the mechanism and series of DOEs. The commands for machine parameter
adjustment will be sent from Matlab or Python to LabVIEW, thus the continuous feeding of the
separator can be assured by this cyber-physical control loop. The sensor network upgrade and
the cyber-physical closed control loop for the entire LIB disassembly system will be completed
in the future development of this project.
66
2.5 Conclusion
An automated disassembly system for recycling end-of-life Z-folded pouch LIBs has
been introduced and prototypically realized. Customized tool sets aiming for automatic material
handling are designed in each module. Success in treating dummy cells proved the effectiveness
of the proposed disassembly strategy. Within this automated disassembly system, cathode
sheets, anode sheets, separators, and Al laminated film housing are automatically separated.
The integrity of cathode sheets can be well preserved. Compared to the destructive crashing
strategy, such improvement will greatly benefit downstream recycling processes, especially for
the direct regeneration strategy. Further research activities will focus on upgrading the vision-
sensor network and the cyber-physical closed loop control system through the entire prototyped
disassembly system so that the reliability and flexibility of the prototyped system can be assured.
67
Chapter 3 Cathode Coating Separation of Lithium Cobalt Oxide
Battery
Recycling EoL LIBs is not only necessary but also urgently needed in recent years. EoL
LIBs are metal enriched city “mine” for lithium, cobalt, manganese and nickel. Generally 1 ton
of lithium can be recycled out of 28 tons of EoL lithium ion battery184,185 which takes 250 tons
of mineral ore spodumene along with 1900 tons of water to extract same amount of lithium186.
Meanwhile, discarded LIBs are serious environmental hazards. Residual electrical capacity
tends to cause explosions or fire accidents. Commonly used LIB electrolyte salt (e.g., Lithium
hexafluorophosphate) reacts with water and release harmful hydrofluoric acid vapor. In terms
of the urgency, the increasing demand of LIBs in EV market since 2010 187 indicates there will
be a heavy burden on OEMs and governments in the foreseen futures since merely 4 to 15 years’
battery module lifetime 173,174 are expected. At the same time, the bursting demand of LIBs will
also impose great pressure on the supply chain of critical raw materials such as cobalt. The
price of cobalt rose by more than 80% over 2017188.
Currently, there are three main recycling methods in industry: pyrometallyrgical
recycling (PR), hydrometallurgical recycling (HR), and direct recycling (DR) 99,176,189. Both PR
and HR methods break the cathode compound down to elemental constituents and selectively
extract metal elements from the mix74. The simplicity of the overall metallurgical process comes
with high energy consumption, large waste generation, and low capability in recovering
Lithium and Manganese83, which is challenging for enterprises to make profit out of the battery
recycling business. In contrast, DR has the highest material recovery rate and least waste
generation among all three recycling methods, and has been actively developing in research
68
labs towards industrial scale applications 80,81,115,177,190. Since the cathode morphology is well-
preserved during the entire direct recycling process, the cathode materials instead of elemental
constituents can be recycled and reused.83
Commonly used materials extraction processes in LIB recycling involve some types of
physical or chemical separation process, such as shredding106, thermal treatment97, and organic
solvent methods. Shredding method, which involves multilevel crushing, fine sieving and air-
classification102, introduces a tremendous amount of impurities that are hard to purify in
subsequent processes. Thermal treatment method easily leads to change in cathode materials
structure, composition and morphology. In our DR, we have developed a novel pre-sorting
process to separate cathode sheet, anode sheet and separator191 and an organic solvent extraction
process to retrieve active cathode powder by dissolving the binder (e.g., Polyvinylidene
Fluoride, PVDF) with the sonication assisted solvent soaking method.
This chapter focuses on the organic solvent method and studies the relationship between
the processing parameters and the cathode materials retrieval yield in the organic solvent
method using Taguchi DoE methods and Regression Analysis. Processing parameters that have
minor influences on the materials retrieval yield from a single cathode sheet are first identified
by Placket-Burman parameter screening method and set to a level that would benefit the yield
the most. The remaining parameters along with essential parameters for the potential mass
production process are evaluated by Taguchi DoE. Finally, the results of Taguchi DoE are used
to generate a regression model that is able to predict the yield under different input parameter
combinations.
69
3.1 Experiment Setup
EoL battery cells (Figure 3-1) used in this work are randomly selected from a pool of
waste drone battery modules provided by a third-party battery recycler. In order to minimize
the residual energy, the battery modules are discharged by the BD200 battery discharger (Figure
3-2 (a)) to an average cell voltage of below 2V/battery. Discharged modules are then manually
disassembled into individual cells.
Figure 3-1 Battery modules disassembled for DOE.
Voltage of each single pouch cell needs to be double checked before being cut open for
extracting the electrode-separator compound (ESC). Z-folded ESC structures are found on all
selected modules, which results in 60 to 70 single sheet electrodes from each battery. Our
previous research191 indicated that fully automated cathodes and anodes separation of Z-folded
Li-ion batteries is feasible for automation, thus we only focus on process parameters that
influence the yield of cathode powder extraction from the cathode sheets. Separated cathode
electrode sheets are then collected for the subsequent soaking and sonicating process as shown
in Figure 3-3. The cathode active materials are retrieved by breaking particle-to-particle and
70
particle-to-Al current collector bonding forces formed by binders. The most commonly found
binder type on a LIB is PVDF which dissolves in organic solvents such as NMP, DMAC, DMF,
DMSO, and acetone. X. Song 81 indicated DMAC and DMF outperform the other solvents on
dissolving effectiveness and cost efficiency. Thus DMAC and DMF are used as soaking and
sonication media in our experiments. The high temperature soaking and sonicating process are
conducted in a convection oven (Figure 3-2 (b)) and ultrasonic cleaner (Figure 3-2 (c)).
Figure 3-2 Equipment requirement (a) Battery module discharger, (b) High temperature oven,
and (c) Ultrasonic cleaner.
The cathode materials yield of the materials retrieval process is estimated by weight
difference of the cathode electrodes before and after the soaking and sonicating process. On
average, the weight of cathode electrodes (𝑊𝑖𝑛𝑖𝑡𝑖𝑎𝑙) consists of 28% Al current collector and
72% cathode coating. After the cathode separation process, residual cathode electrodes are dried
in oven and weighted (𝑊𝑝𝑜𝑠𝑡) again. The final yield of the cathode separation process can then
be estimated by Eq.3-1, which is referred to as response Y and yield in this section.
𝑺𝒆𝒑𝒂𝒓𝒂𝒕𝒊𝒐𝒏 𝒀𝒊𝒆𝒍𝒅% = (𝑾𝒊𝒏𝒊𝒕𝒊𝒂𝒍 − 𝑾𝒑𝒐𝒔𝒕) (𝑾𝒊𝒏𝒊𝒕𝒊𝒂𝒍 ∗ 𝟎. 𝟕𝟐⁄ ) Eq.3-1
71
Figure 3-3 Experiment flow with single cathode electrodes.
3.2 Plackett-Burman Parameter Screening Experiment
3.2.1 Plackett-Burman Parameter Screening Experiment Design
A 5 factor 2 level Plackett-Burman parameter screening experiment is constructed to
identify the control factors that would influence the yield when dealing with a single cathode
sheet. The idea of Placket-Burman experiment design is to ensure each combination of levels
for any pairs of factors are studied for the same number of times, similar to a complete factorial
design but with smaller number of runs. Interactions between the factors are considered
negligible in the Placket-Burman experiment. Other than the aforementioned soaking media,
the soaking process prior to the sonication process also introduces two potential major
quantitative factors: soaking time and soaking temperature. Soaking times are set to be 2h and
6h and soaking temperatures are set to be 60℃ and 90℃. Meanwhile, the sonication process
72
also introduces time and temperature as potential major quantitative factors. In the Placket-
Burman experiment, a single electrode sheet is treated one at a time to prevent interactions
between electrode sheets from influencing the PVDF dissolving and cathode powder escaping.
Thus the sonication time required is relatively short compared to the following Taguchi
experiments and 2 levels of the sonication time are set to 10s and 20s. Two levels of the
sonication temperature are set to 40℃ and 60℃, which is assured by the heating unit integrated
inside the ultrasonic equipment. The ultrasonic equipment available is equipped with a
transducer with a fixed frequency of 40 kHz at 120W power level. Thus a total of 5 factors are
available for the Plackett-Burman experiment design.
For factors and levels given in Table 3-1, a complete factorial design will require 32
runs while Plackett-Burman design only needs 12 runs (Table 3-2). Each run is repeated 5 times
with a single cathode sheet randomly picked from a shuffled electrode pool and responses are
averaged and recorded in the last column of Table 3-2. The averaged responses are further
tested with analysis of variance (ANOVA) to determine the significance of each factor.
ANOVA is an analysis tool used in statistic research that categories input variables inside an
experiment data into two categories: systematic factors and random factors. The systematic
factors have a statistical significant influence on response or the output, while the random
factors will be identified and factored out. The working principle of the ANOVA analysis will
be further explained with the Plackett-Burman experiment data in the following section.
Meanwhile, the ANOVA test is also helpful in a regression study to determine how independent
variables influence the response, which will be demonstrated in the Taguchi experiment later
on.
73
Table 3-1 Screening experiment parameters and levels.
RunOrder Soaking
Media(A)
Soaking Time
(B)
Soaking
Temperature
(C)
Sonicating
Time
(D)
Sonicating
Temperature
(E)
LCO
Separation
Yield (Y)
1 DMAC 3 60 10 40 63.7 2 DMF 1 90 10 40 50.4 3 DMAC 3 90 10 60 70.2 4 DMF 1 60 10 60 52.6 5 DMAC 1 60 10 40 49.5 6 DMF 3 60 20 60 61.7 7 DMF 1 90 20 40 75.2 8 DMAC 3 90 20 40 95.1 9 DMF 3 90 10 60 64.9
10 DMF 3 60 20 40 80.3 11 DMAC 1 60 20 60 93.4 12 DMAC 1 90 20 60 90.2
Table 3-2 Plackett-burman screening experiment design and result.
3.2.2 Parameter Screening Results
The ANOVA analysis result constructed at 95% confidence interval (CI) from Minitab
19 is as shown in Table 3-3. CI is a range of values calculated by statistical methods which
includes the desired true mean with a probability defined in advance. That says, if confidence
intervals are developed with a given CI from an infinite number of independent sample data,
Factor Soaking Media(A) Soaking Time(B) Soaking
Temperature(C) Sonicating Time(D)
Sonicating
Temperature(E)
Unit Type Hour ℃ Second ℃
Level 1 DMAC 2 60 10 40
Level 2 DMF 6 90 20 60
74
the frequency of these intervals containing the true mean equals to the given CI192,193. For
example a 95% CI is a range of values that you can be 95% certain contains the true mean of
the population (or other population parameters). CI is designated before examining the
experiment data and 90% and 99% CIs are also frequently used in analysis. P-value in the last
column of Table 3-3 is a probability that reflects the measure of evidence against the null
hypothesis between the corresponding independent input parameters and the dependent output
response. A smaller p-values correspond to stronger evidence. If the p-value is below (1 − CI),
which equals 0.05 in this study, then the null hypothesis is rejected and the corresponding
independent input parameter is designated as “statistically significant”194,195. Columns of the
ANOVA table for the Plackett-Burman experiments are source, degree of freedom (DF),
adjusted sums of squares (Adj SS), adjusted mean squares (Adj MS), F-value, and P-value from
left to right. The DF represents the number of information in the data. The ANOVA analysis
uses this information to figure out the values of unknown population parameters. An
independent variable with 𝑘 level will always have 𝑘 − 1 DF in order to get the number of
values/levels that are free to vary in a data set. The Adj SS are measures of variation for different
variables of the model. In the ANOVA table, Adj SS is divided into Term, Error, and Total.
Minitab uses the Adj SS for term to calculate the P-value for a variable. Adj MS measures how
much variation a variable or a source explains, thus DF of a term is taken into consideration. F-
value is the test statistic used to determine whether the independent variable is associated with
the response. A sufficiently large F-value indicates that the independent variable is significant.
P-value can then be calculated from DF of the specific input variable, DF of the Error, Adj SS,
Adj MS, and F-value within an F-distribution form at a CI of 95%.
75
As aforementioned, input variables with P-values less than 0.05 proved to have enough
evidence against the null hypothesis, thus considered to have statistically significant
contribution to the response which is the material separation rate in this case. The sonicating
time with P-value in 0.003 proved to be a decisive factor towards the material separation rate
and will be factored in for the following Taguchi experiment design. Figure 3-4 is the main
effect plot of each input variable at different levels. It displays the means for each level within
a variable and connects these points with a broken line. If the line is parallel to the x-axis, there
is no main effect present. The response mean remains the same across all factor levels. If the
line is not horizontal, the steeper the slope of the line, the greater the magnitude of the main
effect. As a quantitative factor, the soaking media have the p-value of 0.046 and the main effect
plot in Figure 3-4 shows DMAC outperforms DMF, thus DMAC is chosen to be the organic
solvent media to be applied in Taguchi experiment design. The P-value of other factors from
Table 3-1 are higher than 0.05, thus resulting in much flat main effect plots. These factors agree
with the null hypothesis and are taken out of the factor list for the following Taguchi experiment
design and their pre-set levels are listed in Table 3-5 along with other essential pre-set process
parameters.
Source DF Adj SS Adj MS F-Value P-Value
A 1 494.08 494.08 6.27 0.046
B 1 50.43 50.43 0.64 0.454
C 1 167.25 167.25 2.12 0.195
D 1 1742.43 1742.43 22.12 0.003
E 1 29.45 29.45 0.37 0.563
Error 6 472.57 78.76
Total 11 2956.22
Table 3-3 Analysis of variance of screening experiment.
76
Figure 3-4 Main effects plot for fitted means of screening experiment yield.
3.3 Taguchi DOE
Studying one single electrode at a time can be done at lab-scale to better understand the
relationship between processing parameters and materials retrieval yield in organic solvent
extraction process. In an industrial relevant environment, however, mass production requires
much higher process throughput and introduces more process parameters that need to be
systematically studied. Thus to simulate the industrial materials retrieval process, shuffled
cathode electrodes are cut into smaller pieces with a controlled size and treated at a much higher
solid-to-liquid weight ratio in the following study.
3.3.1 Experiment Design
3 factors with 4 levels from each factor are selected as the independent variables for the
Taguchi experiment. Sonication time has been proved to be statistically significant to influence
the cathode coating separation efficiency when dealing with the single electrode in the Plackett-
Burman experiment. Preliminary experiment indicates that a sonication time on minute scale
instead of seconds is required for getting a satisfactory separation yield when dealing with
interactions between cathode electrodes and at least tripled solid-to-liquid ratio. Hence, 4 levels
77
of the sonication time are set to 1 min, 3 min, 5 min, and 7 min accordingly. The other
independent variable introduced by cutting the cathode electrode into smaller pieces is the sheet
size. The largest sheet size involved in the Taguchi DOE is 4.18cm2, which is a quarter of the
original intact sheet size. The rest three levels of the sheet sizes are downsized in the order of
0.5 time. The final independent variable is the solid-to-liquid weight ratio. The lowest solid-to-
liquid weight ratio introduced in the Taguchi DOE is 5mg/ml, which is already three times of
the initial solid-to-liquid weight ratio in the Plackett-Burman experiment. The rest three levels
are sequentially doubled and the solid-to-liquid weight ratio reaches 20mg/ml at level 4.
Taguchi DOE is utilized to identify the significance of each control factor chosen in
Table 3-4. A regression model is also developed based on Taguchi DOE to predict the response
according to levels of independent variables. The Taguchi DoE method is an effective
statistical off-line quality control methodology aiming at increasing the robustness of a product
or process facing variations over which we have minimum control in the design stage. It is
capable of studying both control factors that are controllable during the production, and noise
factors that we cannot control when the process is in use. Experiment designed by Taguchi
method is able to identify controllable factors (independent variables) that minimize the effect
of the noise factors. In a complete Taguchi experiment, noise factors are manipulated so that
variability occur on purpose. The optimal control factor (independent variables) can then be
determined to improve the robustness of the production process or the product itself by
increasing the system resistance to variations from the noise factors. Here in this study, the
Taguchi DOE is applied mainly due to its superior capability of dealing with multi-level
variables with its unique Orthogonal Array in variable combination subset selection. For a
complete n factors and m levels factorial design, a total number of mn times of repeating
78
experiments are needed. In the case of 3 factors and 4 levels, the original 64 times repeating
experiments can be decreased to 16 times with minimum information loss. The efficiency will
dramatically increase as the number of factors and levels increase. Details about Orthogonal
Array selection have been explained in many publications196-198.
Table 3-4 lists all 3 factors and 4 levels to be studied in Taguchi DoE, which enables us
to develop a more precise regression model compared to the 2 level Placket-Burman parameter
screening experiment. In this study, the L16 Taguchi orthogonal array is applied to select
subsets from combinations of 3 control factors at 4 levels. The Taguchi orthogonal arrays are
well balanced to independently evaluate all levels and factors thus each factors and levels can
be equally considered. The detailed subset combinations are as listed in Table 3-6. Each subset
combination in L16 Taguchi experiment design is repeated for 5 times and yields are averaged
and recorded in the last column of Table 3-6 Taguchi L16 orthogonal array. The statistical
analysis of yields utilizing ANOVA are carried out using MINITAB.
Table 3-4 Taguchi experiment parameters and levels.
Table 3-5 Pre-set process parameters.
Factor Sonicating Time(A) Sheet Size(B) Solid-liquid Weight Ratio(C)
Unit min cm^2 mg/ml
Level 1 1 0.52 5
Level 2 3 1.04 10
Level 3 5 2.09 15
Level 4 7 4.18 20
Parameters Soaking Media Soaking Time Soaking
Temperature
Sonicating
Temperature
Sonicating
Frequency
Unit NA Hour ℃ ℃ kHz
Level DMAC 5 90 60 40
79
RunOrder Sonication time(A) Sheet Size(B) Solid-liquid Ratio(C) Separation Yield(Y)
1 1 0.52312 5 25.4
2 1 1.04625 10 23.5
3 1 2.09250 15 21.9
4 1 4.18500 20 23.0
5 3 0.52312 10 30.8
6 3 1.04625 5 36.4
7 3 2.09250 20 31.2
8 3 4.18500 15 39.7
9 5 0.52312 15 37.8
10 5 1.04625 20 34.2
11 5 2.09250 5 57.9
12 5 4.18500 10 51.7
13 7 0.52312 50 52.3
14 7 1.04625 15 60.9
15 7 2.09250 10 69.8
16 7 4.18500 5 79.9
Table 3-6 Taguchi L16 orthogonal array and the corresponding separation yield.
3.3.2 Taguchi DOE Results
A continuous quality loss function is used to evaluate the performance characteristics
of the Taguchi DoE. This loss function calculates the deviation of a design parameter from the
desired value. Value of this loss function is called the signal-to-noise (S/N) ratio. Three
categories of S/N ratios are available depending on the goal of experiments:
If the response is to be maximized, that is the larger is better, then:
𝑆 𝑁⁄ = −10 ∗ 𝑙𝑜𝑔 (∑(1 𝑦𝑖2⁄ )/𝑛)
80
If the response is better to be an intermediate value, then:
𝑆 𝑁⁄ = −10 ∗ log (∑(�̅�2/𝜎2))
If the response is to be minimized, that is the smaller is better, then:
𝑆 𝑁⁄ = −10 ∗ log (∑(𝑦𝑖2)/𝑛)
where 𝑦𝑖 is the response of the 𝑖𝑡ℎ run, �̅� is the average response of all runs, σ is the standard
deviation of the response, and n is the total number of runs.
Here in our case, to study the yield of the cathode material retrieval, we expect the
response to be the higher the better. Thus S/N ratios for yield are calculated from the larger is
better S/N calculation equation. For the next phase of the study, the aluminum impurity will be
introduced as the second response. S/N ratios for this response will be calculated by “the smaller
is better” principle as aluminum debris from the current collector is an unwanted element in the
cathode materials retrieved.
The processing parameters rank towards their influence on the yield as well as how
parameters influence the yield are shown in the response table for S/N ratio (Table 3-7) and
main effect plot (Figure 3-5). The parameters that influence the cathode yield from high to low
are sonication time (A), solid-liquid weight ratio(C), and sheet size (B). Both sonication time
and sheet size have positive impact on the yield as their level increases, while the solid-liquid
weight ratio shows negative influence on the yield as more electrodes being added into certain
amount of organic solvent.
81
Level Sonication time(A) Sheet Size(B) Solid-liquid Ratio(C)
1 27.39 30.95 33.16
2 30.71 31.25 32.08
3 32.94 32.21 31.51
4 36.25 32.88 30.54
Delta 8.86 1.94 2.61
Rank 1 3 2
Table 3-7 Response table for S/N ratio.
Figure 3-5 Main effects plot for S/N ratio.
To quantify the importance of each control factors and their specific contribution
towards the yield, an analysis of variance of the S/N ratio is carried out with a confidence
interval of 95%. This means that as long as the P-value of a factor shown in Table 3-4 is less
than 0.05, this control factor can be considered to have statistically significant influence on the
cathode yield. Individual contribution(𝑃%) of the 𝑖𝑡ℎ factor towards the cathode yield is
calculated by Eq.3-2 and recorded in the last column of Table 3-8.
𝑷𝒊% = 𝑺𝒆𝒒 𝑺𝑺𝒊 𝑺𝒆𝒒 𝑺𝑺𝒕𝒐𝒕𝒂𝒍⁄ Eq.3-2
82
The P-value of the sonication time (A), the sheet size(B), and the solid-liquid weight
ratio(C) proved to be well below 0.05 and the P-value rank agrees with the rank indicated by
the response table for S/N ratio. The calculated individual contribution (𝑃%) in Table 3-8
indicate that the sonication time has the highest contribution of 86.55% as followed by solid-
liquid weight ratio at 7.44% and sheet size at 4.90%.
Table 3-8 Taguchi experiment analysis of variance for S/N ratio.
3.3.3 Linear Regression Model for Yield Prediction
As all three factors proved to be statistically significant toward the cathode separation
rate, the following linear regression model is established to predict the yield from levels of all
factors by analyzing results of Taguchi experiments. Linear regression model is an equation
that minimizes the distance between the fitted line and all experiment data. Generally speaking
a linear regression model fits the experiment data well if the distances between the predicted
value and the observed values are unbiased. To statistically determine how well the model fits
the original data, three goodness-of-fit statistics need to be examined in the summary below the
developed linear model: S, R-squared (R-sq), and R-sq (adj). S is used to assess how well the
linear model describes the response. It represents how far the true value falls from the value
predicted by the linear model. The lower the value of S, the better the model describes the
Source DF Seq SS Adj SS Adj MS F-Value P-Value P%
Contribution
A 3 166.807 166.807 55.6023 155.43 0.000 86.55
B 3 9.443 9.443 3.1478 8.8 0.013 4.90
C 3 14.343 14.343 4.7810 13.36 0.005 7.44
Error 6 2.146 2.146 0.3577 1.11
Total 15 192.74 100
83
response. R-sq also measures how close the experiment data points are to the regression line,
but it is defined by the percentage of the response variable variation that is explained. R-sq is
always between 0% and 100% and the higher the R-sq, the better the linear model fits the
experiment result. However, R-sq always increases when more terms are added to the model,
even if no actual improvement has been added. Thus R-sq (adj) is introduced to compare models
that have different numbers or terms. It can incorporate the number of terms to help us choose
the best fit model. From the summary table, both R-sq and R-sq(adj) are at a level higher than
95%, which indicate the developed linear regression model is a good fit to the Taguchi
experiment result.
𝑳𝑪𝑶 𝑺𝒆𝒑𝒂𝒓𝒂𝒕𝒊𝒐𝒏 𝒀𝒊𝒆𝒍𝒅 = 𝟐𝟎. 𝟑 + 𝟔. 𝟖𝟗 ∗ 𝑺𝒐𝒏𝒊𝒄𝒂𝒕𝒊𝒏𝒈 𝑻𝒊𝒎𝒆 + 𝟑. 𝟐𝟖 ∗ 𝑺𝒉𝒆𝒆𝒕 𝑺𝒊𝒛𝒆 −
𝟎. 𝟗𝟔 ∗ 𝑺𝒐𝒍𝒊𝒅/𝑳𝒊𝒒𝒖𝒊𝒅 𝑹𝒂𝒕𝒊𝒐 Eq.3-3
Term S R-sq R-sq(adj)
Value 3.71270 96.52% 95.65%
Table 3- 9 Summary table of the regression model for LCO separation yield.
To validate the accuracy of the linear regression model, confirmation tests are requested
to compare the predicted cathode separation rate and the experiment value. As shown in Table
3-10, three additional runs are randomly picked from the factor combination pool of the full
factorial design. The averaged error of three validation experiments is 3.35%, which is within
an acceptable range.
84
Experiment
No.
Sonication
Time Sheet Size
Solid-liquid
Ratio
Predicted
Value (%)
Experiment
Value (%) Error (%)
1 3 0.523 15 28.26 26.75 5.36
2 5 1.485 10 58.58 58.99 0.22
3 7 2.092 5 70.56 73.73 4.49
Table 3-10 Linear regression model for predicting LCO separation yield verification test result.
3.4 Conclusion
In summary, this study has identified sonicating time, sheet size, and solid-liquid weight
ratio as three essential control factors towards the efficiency of the pre-treatment process,
especially the organic solvent step. DMAC outperformed DMF and proved to be the most cost
effective organic solvent with the highest efficiency for the task. The S/N ratio analysis in
Taguchi DOE revealed the contribution of sonicating time (86.55%), sheet size (7.44%), and
sheet size (4.90%) towards the final yield. The mathematical relationship between the yield and
control factors were successfully established with the result of Taguchi DOE and proved
accurate by confirmation tests. The developed linear regression model provide a reference in
choosing the proper control factors/independent variables with satisfying accuracy for the
future industrial scale production. The success in the yield prediction enables us to study
aluminum impurity introduced by sonicating process as the second response for the next step.
85
Chapter 4 Cathode Coating Separation of Lithium Iron Phosphate
Battery
Previous study on predicting the organic solvent approach separation yield utilizing
DOE and regression analysis has revealed that a precision linear regression model is able to
link the yield with multiple independent input process parameters for EoL LCO cells. However,
LFP cell is also a strong competitor in the LIB market, especially the EV market in China.
Among the 15.7 GWh sales volume of power batteries in China from 2015, LFP rechargeable
batteries accounted for 69%199. Though the energy density and operating voltage of LFP are
lower than that of LCO chemistry, LFP cells hold inherent advantages in low cost, long-term
stability, low toxicity, and well-defined performance. It can be foreseen that, in the short future,
EoL LFP EV battery packs will rapidly accumulate and pose a pressure on the full product cycle
fulfillment. Hence, the yield of treating EoL LFP utilizing organic solvent approach is studied
with, again, the DOE method, but with full-factorial experiment design. Since the preliminary
study for this Chapter reveals that the organic solvent approach might as well avoid introducing
the sonication bath to facilitate the separation of cathode coating and the Al current collector,
the number of the parameter drops to a scale where full-factorial experiment is practical. Similar
to the LCO study conducted in Chapter 3, a regression model is built to predict the newly
defined process yield from levels of independent input variables. An additional yield prediction
approaching applying Latin square with the limited number of experiments needed is also
introduced at the end of this Chapter.
4.1 Full-Factorial Experiment Setup
Tenergy 3.2V 2500mAh LFP (IFR26650P) rechargeable batteries as shown in Figure
4-1(a) are chosen as the test subject for this study. 26650 in their configuration code represent
86
a shape standard of 26.5 mm in diameter and 65.4mm in length similar to the popular cylindrical
18650 and 21700 cells. The cylindrical cell is one of the most widely adopted packing styles
for secondary batteries. The simple structured ESC that produced from the circular winding
process allows the cylindrical cells to better fit the automation environment, thus achieving
higher production throughput and a lower cost-per-kWh. The fact that continuous electrodes
and separators do not need any cutting step in the ESC production process also contribute to
the minimization of production cost by creating a better product consistency. The cases of
cylindrical cells are formed from either steel plate or aluminum plate, which provide solid
protection to the ESC against the external impact as well as helping with the heat dissipation.
Though the uniquely designed gas releasing vent as shown in Figure 4-1(b) increases the
assembly difficulty, it provides the incomparable advantage in gas venting and pressure
releasing. Thus the housing of the cylindrical cell is finally sealed directly after the electrolyte
filling process, without needing an extra gas releasing step in the formation process as the
prismatic pouch cells.
To imitate the condition of EoL, IFR26650P LFP cells are cycled on a battery cycler
under 2C rate between a voltage range from 2.8V to 3.6V until their capacity falls below 85%
of their nominal capacity. The continuous cathode electrodes adopted in IFR26650P are
measured to be 52mm in width and 0.8m in length. Manually extracted electrodes are then
evenly cut into smaller pieces along the length, which result in 52mm*40mm size cathode
electrodes for each experiment. NMP is chosen to be the soaking media as preliminary
experiments have revealed an exceedingly higher efficiency of NMP compared to the other
organic solvent candidates such as DMAC, DMF, or DMSO in this proposed organic solvent
approach.
87
Figure 4-1 (a)Tenergy 3.2V 2500mAh LFP (IFR26650P) power cell rechargeable battery and
(b) Typical internal structure of a cylindrical LIB.
Two independent input variables are soaking time and soaking temperature, which
proved to be sufficient for developing a precision regression model. Each variable is studied at
four levels as shown in Table 4-1. Soaking time is studied at 1h, 2h, 3h, and 4h and soaking
temperature is studied at 80℃, 85℃, 90℃, and 95℃. Thus this 2 factors and 4 levels full
factorial experiment results in 16 combinations in total.
Factors Response
A B Y
Name Soaking Time Soaking
Temperature
Separation Yield
Unit Hour ℃ %
Level 1 1 80 NA
Level 2 2 85 NA
Level 3 3 90 NA
Level 4 4 95 NA
Table 4-1 Full-factorial experiment parameters and levels.
The original weight of a 52mm*40mm cathode electrodes is around 680mg among
which around 550mg is cathode LFP coating and the rest 130mg is the Al current collector.
After being soaked in NMP under high temperature for hours, a unique winkle structure will
88
appear as the adhesion between the cathode coating and the current collector gets partially
destroyed as shown in Figure 4-2. As the soaking time gets longer, the wrinkle area will expand
as shown in Figure 4-2 (b), (c), and (d) sequentially. The winkled cathode coating can then be
easily peeled off almost naturally from the cathode electrodes. The net weight of the remaining
cathode electrode (𝑊𝑟𝑒𝑚𝑎𝑖𝑛) will be measured again after being sufficiently dried in the oven.
The final separation yield that represents the percentage of cathode coating that is “soaked off”
from the cathode electrode can then be calculated with the following equation.
𝑺𝒂𝒑𝒂𝒓𝒂𝒕𝒊𝒐𝒏 𝒀𝒊𝒆𝒍𝒅 % = (𝟔𝟖𝟎 − 𝑾𝒓𝒆𝒎𝒂𝒊𝒏) 𝟓𝟓𝟎⁄ Eq.4-1
Figure 4-2 The appearance of (a) the original LFP cathode electrode and the appearance of the
LFP cathode electrode soaked in NMP under 90℃ after (b) 2h, (c) 3h, and (d) 4h.
89
4.2 Result and Discussion
4.2.1 Full-factorial DOE Separation Yield
The separation yield of each variable combination is calculated with the aforementioned
approach and yield of each combination is averaged between three repeated experiments. Thus
a total of 48 pieces of 52mm*40mm cathode electrodes are needed, which requires continuous
cathode electrodes from 3 EoL IFR26650P LFP cells. After manually dismantle these cells,
continuous cathode electrodes are evenly cut into 52mm*40mm pieces. These cathode electrode
pieces are then fully shuffled and randomly picked cathode electrode pieces from this pool are
to be assigned to each experiment later on. Such precaution minimizes the influence of possible
noise factors introduced from inconsistency between different cells and electrode pieces. The
averaged separation yield of all 16 combinations are as shown in Table 4-2. From Table 4-2, it
is obvious that both independent variables are positively correlated with separation yield. The
main effect plot as shown in Figure 4-3 indicates a similar rising trend between sonication time
and sonication temperature in the selected range, which confirms the positive correlation
between the separation yield and both independent variables.
Soaking Temperature
S
oak
ing T
ime
Level 1 Level 2 Level 3 Level 4
Level 1 0 0 1% 8%
Level 2 0 1.2% 14% 30%
Level 3 2.0% 6.7% 56.5% 90%
Level 4 5.6% 33.6% 83.9% 95%
Table 4-2 Separation yield of the cathode coating in 2 factors, 4 level full-factorial experiment.
90
Figure 4-3 Main effect plot for separation yield.
From the post analysis of this experiment series, a unique discovery worth mentioning
is the great potential that the organic solvent approach shows in recovering LFP cathode coating
without introducing any brutal vibration or mechanical forces to separate cathode coating and
the electrode. The sample as shown in Figure 4-4 (a) is soaked in NMP for 4.5h under 95℃,
which is slightly longer compared to the sample soaked for 4h under 95℃ that achieved 95%
separation yield. It is clear that LFP cathode coating material originally adhered on both sides
of the current collector naturally detached as whole pieces, which leaves a smooth Al current
collector separated from the cathode coating without involving any extra actions other than
high-temperature soaking. Though adhesion between particles on cathode coating sheets is
considerably weak after NMP dissolves the majority of PVDF, the intact cathode coating sheets
are extracted from NMP solvent with caution. The extracted cathode coating sheets are then
tiled on a weighing paper with the 52mm*40mm Al current collector covered on top as shown
in Figure 4-4 (b). The surface area of the cathode coating sheet is apparently larger than the Al
current collector that it is originally coated on, which indicates that the volume of the cathode
coating sheet is expanded during the NMP soaking process. The volume expansion might as
well be the explanation to the appearance of winked structures on the electrode surface shown
91
in Figure 4-2. In order to quantify the volume expansion of the cathode coating sheet, 10 LFP
cathode electrode pieces randomly picked from the aforementioned pool are soaked inside the
NMP under high temperature until cathode coating sheets fully detach from the Al current
collector. The averaged surface area of the cathode coating sheets is 23.96cm2, which is a 15.2%
increase compared to the 52mm*40mm cathode electrode pieces with 20.80cm2 surface area.
Since the thickness of the cathode coating sheet remains at around 150μm before and after the
NMP soaking process, the total volume expansion of the cathode coating sheet can be
considered the same as the surface area expansion. Though the true mechanism for the volume
expansion phenomenon requires further investigation, the proposed NMP soaking process
indicates a great potential in separating and recollecting the cathode coating from EoL LFP
cells with a 100% purity even for industrial-scale production.
Figure 4-4 Fully separated LFP cathode coating (a) naturally detached from the Al current
collector during soaking, and (b) coating expansion comparison with the Al current collector.
92
4.2.2 Regression Analysis
The regression model developed based on the separation yield from Table 4-2 is as
shown in Eq.4-2. A total of four terms are included in the model. Other than the expected terms
of constant, soaking time, and soaking temperature, the fourth term proved to be Soaking Time ∗
Soaking Temp, which indicate a nonlinear relationship between two independent variables and
the separation yield. The R-sq and R-sq(adj) from the summary table of the regression model
are both above 90%, which indicate a fair fit of the model to the experiment observations.
𝑳𝑭𝑷 𝑺𝒆𝒑𝒂𝒓𝒂𝒕𝒊𝒐𝒏 𝒀𝒊𝒆𝒍𝒅 = 𝟏𝟐𝟔 − 𝟏𝟕𝟐. 𝟔 ∗ 𝑺𝒐𝒂𝒌𝒊𝒏𝒈 𝑻𝒊𝒎𝒆 − 𝟏. 𝟔𝟔 𝑺𝒐𝒂𝒄𝒌𝒊𝒏𝒈 𝑻𝒆𝒎𝒑 +
𝟐. 𝟏𝟖𝟑 𝑺𝒐𝒂𝒌𝒊𝒏𝒈 𝑻𝒊𝒎𝒆 ∗ 𝑺𝒐𝒂𝒌𝒊𝒏𝒈 𝑻𝒆𝒎𝒑 Eq.4-2
Term S R-sq R-sq(adj)
Value 10.4498 92.84% 91.05%
Table 4-3 Summary table of the regression model for LFP separation yield.
The ANOVA table for the regression model analysis the significant level of each term
appears in Eq.4-2. Noticing that the DF of each term is considered as 1 no matter how many
levels the term originally possesses. The total DF equals the total number of observations minus
1, thus in this case equals to 15. The DF of the error then equals the total DF minus the total
number of terms involved in the model. The P-value of the soaking time term equals 0.001,
which is way less than 0.05. Thus the soaking time term proved to be statistically significant to
the LFP separation yield. The P-value of the soaking temperature term equals 0.173, which
indicates that the soaking temperature term alone has a much less significant influence towards
the separation yield compared to the soaking time term. However, the multiplication of the
soaking time and the soaking temperature, which is the only nonlinear term in the regression
model, proved to be the most statistically important term to the LFP separation yield with a P-
93
value even less than 0.001. Next, the accuracy of the developed model is testified by three
supplementary experiments.
Source DF Adj SS Adj MS F-Value P-Value
A 1 2421.5 2421.5 22.17 0.001
B 1 228.8 228.8 2.10 0.173
A*B 1 2978.4 2978.4 27.08 0.000
Error 12 1310.4 109.2
Total 15 18305.8
Table 4-4 ANOVA table for the regression analysis.
The verification of the linear regression model developed in Chapter 3 randomly picked
three supplementary experiments from the factor combination pool of the full factorial design
and calculated the error of the LCO separation yield from input variables. However, the ultimate
goal of the discovered LFP soaking process is to achieve the 100% separation efficiency with
the minimum energy consumption and the highest time efficiency. Thus a better approach to
testify the accuracy of the developed nonlinear regression model is to predict the minimum
soaking time needed to achieve 100% LFP separation yield under preselected levels of the
soaking temperatures. Three levels of temperature, 85℃, 100℃, and 110℃, are selected as
shown in Table 4-5. Status of the soaked samples are checked every 15min for confirming
whether the 100% separation yield has been achieved or not. The averaged error between
predicted minimum soaking time and the observed soaking time for three temperature levels
are around 6.8%, which is within an acceptable range.
94
Experiment No.
Soaking Temperature/℃
Separation Yield
Predicted Soaking Time /h
Observed Soaking Time/h
Error (%)
1 85 100% 8.88 9.75 9.7
2 100 100% 3.01 3.25 7.9
3 110 100% 2.32 2.25 3.0
Table 4-5 Result of the non-linear regression model verification tests.
4.2.3 Correlation Pre-confirm with Contour Plot and Latin Square
As proved in section 4.3.2, utilizing a full factorial experiment design for regression
model training can achieve a decent accuracy towards the prediction of the LFP separation yield.
A total of 16 combinations of the input variables are expected out of the 2 factors and 4 levels
full-factorial experiment. With each combination repeated for 3 times to minimize the noise
variation, a total of 48 observations are conducted and it would have been a total waste of time
and effort if the LFP separation yield does not have anything to do with the two selected input
variables in the first place. Thus in order to pre-confirm the correlation between the selected
input variables and the LFP separation yield with minimum efforts, an approach combining the
Latin Square Design (LSD) and the contour plot visualization approach is conducted prior to
the full factorial experiments designed in section 4.3.1.
The LSD is a very efficient experiment design method when only two input variables
are involved regardless of how many levels they have. In other words, the LSD is usually used
to simultaneously control two sources of variability just as what is needed in the LFP separation
yield study. The LSD gets the name due to the fact that it can be written as a square with Latin
letters to correspond to observing sampling subsets. The number of rows and columns
corresponds to the number of levels of each input variable. Thus, for the 2 factors and 4 levels
95
situation in this study, the Latin Square consists of 4 rows and 4 columns with 4 choices of
sampling subsets, A, B, C, and D, as shown in Table 4-6. Noticing that Table 4-6 is not the only
option for dividing the sampling subsets from a 4 by 4 LSD. The sampling subset design is
considered as a proper LSD as long as each sampling subset only appears once in each row and
in each column. Also because of this restriction, the sampling subsets are orthogonal. Here the
sampling subset “A” as encircled in Table 4-6 is selected as the correlation pre-confirm
experiment and detailed combination is as shown in Table 4-7.
Level
1
Level
2
Level
3
Level
4
Level
1 C A D B
Level
2 D B A C
Level
3 A C B D
Level
4 B D C A
Table 4-6 A LSD of the 2 factors and 4 levels experiment.
Soaking Temperature
S
oak
ing T
ime
Level
1
Level
2
Level
3
Level
4
Level
1
Level
2
Level
3
Level
4
Table 4-7 Latin square design of 2 factors and 4 levels full-factorial experiment.
Similar to the full-factorial experiment introduced in section 4.2.1, each combination
from the LSD is repeated for 3 times. The LFP separation yield is averaged and recorded in
96
Table 4-8 accordingly. The rectangular contour plot can then be plotted with the soaking
temperature as the x-axis, the soaking time as the y-axis, and the LFP separation yield as the z
value. A rectangular contour plot is a graphical technique for representing a 3-D surface by
plotting z value, also named as contours, on a 2-D plane. Lines are drawn for connecting the (x,
y) coordinates where the same z value occurs. The plot can also reveal the relationship between
three explanatory variables and a response variable by the polar contour plot in the 3-
dimensional space or the ternary contour plot in the 2-dimensional plane. In this study, the
rectangular contour plot as shown in Figure 4-5(a) is adopted since only two input variables are
involved. The infill color of the area between contours grows darker as the LFP separation yield
gets higher. With merely 4 sampling subset, a quarter of the full-factorial experiments required,
clear positive correlations between both soaking time/soaking temperature and the LFP
separation yield are revealed by the rectangular contour plot. The full-factorial experiments
aforementioned in section 4.2.1 are then conducted with the confidence that a regression model
can be successfully developed. The rectangular contour plot of the full-factorial experiments as
shown in Figure 4-5 (b) is also plotted with data from Table 4-2. A much similar trend can be
observed between two rectangular contour plots.
Soaking Time/h Soaking Temperature/℃ LFP Separation Yield
3 80 0
1 85 5
2 90 12
4 95 97
Table 4- 8 LFP separation yield of the selected Latin Square sampling subset.
97
Figure 4-5 The contour plot plotted from the result of (a) the full factorial experiments and
(b)the latin square sampling subset.
4.3 Conclusion
The intact LFP coating separation process studied in this chapter offered a great
opportunity in achieving 100% LFP active material reclaiming with minimum process
complexity for industrial-scale recycling of EoL LFP batteries. Compared to the traditional
organic solvent extraction process assisted by the ultrasonic bath introduced in Chapter 3, the
single step soaking process can save much effort in liquid-powder separation and organic
98
solvent reclaim. The success in the development of the LFP separation yield regression model
has everything to do with the volume expansion of the LFP cathode coating material, though
the mechanism behind this phenomenon requires further study. Meanwhile, the future
development of this study should expand the rectangular contour plot to cover a broader range
of the soaking time and the soaking temperature.
99
Chapter 5 LFP Direct Recycling
As aforementioned in Chapter 1, direct recycling process is by far the most
environmental friendly recycling approach with the highest material recovery rate. The
automated disassembly attempt as well as the cathode coating separation study both serve as
the pretreatment steps for the direct recycling process which requires pure cathode coating as
the raw material for the resynthesis of new battery-grid cathode powder. Processes such as heat
sintering, co-precipitation, and sol-gel123 are major resynthesize methods, among which the heat
sintering method involves the direct solid-phase calcination is suitable for treating EoL cathode
materials with simple elemental composition like LCO and LFP. This Chapter focuses on the
regeneration of EoL LFP reclaimed from experiments in Chapter 4 utilizing the heat sintering
method. Samples sintered under different temperatures and atmospheres are first characterized
by SEM and XRD. The rate performance of the regenerated samples are then tested in the form
of coin cells utilizing lithium chips as the counter electrodes.
5.1 EoL LFP Direct Regeneration
The separated LFP cathode coating sheets collected from the experiments conducted in
Chapter 4 are grinded and sieved through a sieve with 325 × 325 mesh size. Then, the lithium
deficiency of the sieved LFP powder caused by battery aging and incomplete discharge then
need to be quantified by the Inductively Coupled Plasma Atomic Emission Spectrometer (ICP-
AES) test which is an analytical technique used for the detection of chemical elements. It is an
emission spectrophotometric technique, exploiting the fact that excited electrons emit energy
at a given wavelength as they return to ground state after excitation by high temperature Argon
Plasma.200 The mole ratio of remaining Lithium and phosphorus (Li/P ratio) is measured by the
ICP-AES to confirm the value of x in the reclaimed Li1−xFePO4 powder. Since ICP-AES
100
analysis requires a sample to be in solution, the 3mg of LFP powder sample is first mixed into
the acid solution consists of 7ml Hydrochloric Acid (HCl), 2ml Hydrogen Peroxide (H2O2), and
6ml DI water. The solution is then sealed and heated in the convection oven under 70℃ for 24h
to facilitate the dissolution. After further diluting the acid solution to 0.467 w%, dissolved
samples are analyzed by ICP-AES provided by Soil Lab in Virginia Tech. The result shows that
the Li/P ratio in the reclaimed Li1−xFePO4 powder is around 0.8, which indicates a 20%
Lithium loss.
Figure 5-1 demonstrates the process flow of the direct regeneration process conducted.
As the ICP result revealed the formula of the EoL LFP to be Li0.8FePO4, the EoL Li0.8FePO4
powders is mixed with LiOH at a mole ratios of LiOH/ Li0.8FePO4 at 20%. Then the
homogeneous mixture powders are pressed into pellets with 0.75 inches in diameter. Four
samples are sintered at 550℃, 650℃, and 750℃ in Ar&H2 atmosphere (abbreviated as 550-
Ar&H2, 650-Ar&H2 and 750-Ar&H2) and 750℃ in Ar atmosphere (abbreviated as 750-Ar).
During the sintering process, Li1−xFePO4 react with LiOH following Eq.5-1.The tube furnace
(OTF-1200X, MTI Corporation) is utilized as the sintering equipment to assure a neutral or
reducing atmosphere. Unlike most LCO/NCM direct regeneration processes that sinter the
mixture in air201-205, the LFP cathode would easily oxidize ferrous for ferric with the presence
of air or oxygen. Jie.et all analyzed the thermal decomposition of the EoL LFP cathode in the
presence of oxygen using Thermogravimetry&Differential Scanning Calorimetry (TG-DSC),
XRD, SEM, and Energy Dispersive Spectrometer (EDS). It is suggested that the oxidation
reaction of iron phosphate as shown in Eq.5-2 occurs with the presence of oxygen during the
high temperature treatment206. Thus in this study three EoL LFP samples are sintered in Ar&H2
atmosphere and one EoL LFP sample is sintered in Ar atmosphere as the control experiment.
101
Pelletized material will then be crushed into fine powders after the heat sintering process for
further tests.
𝑳𝒊𝟏−𝒙𝑭𝒆𝑷𝑶𝟒 + 𝒙𝑳𝒊𝑶𝑯 → 𝑳𝒊𝑭𝒆𝑷𝑶𝟒 + 𝒙 𝟐⁄ 𝑶𝟐 + 𝒙 𝟐⁄ 𝑯𝟐 Eq.5-1
𝟏𝟐𝑳𝒊𝑭𝒆𝑷𝑶𝟒 + 𝟑𝑶𝟐 → 𝟒𝑳𝒊𝟑𝑭𝒆𝟐(𝑷𝑶𝟒)𝟑 + 𝟐𝑭𝒆𝟐𝑶𝟑 Eq.5-2
Figure 5-1 Process flow of the EoL LFP regeneration process.
Property of the sintered samples are then characterized by XRD which is a powerful
nondestructive technique for characterizing crystalline materials. It provides information on
structures, phases, preferred crystal orientations (texture), and other structural parameters, such
as average grain size, crystallinity, strain, and crystal defects. The XRD patterns of products
are measured by a bench-top X-ray diffractometer (Bruker, D2 PHASER). The scanning rate is
to 0.06°/s and the 2θ range is from 15° to 80° (Cu Kα radiation, 40kV, 30mA, λ = 1.5418 Å).
Figure 5-2 shows the comparison of the XRD pattern between six test subjects. XRD peaks in
102
the plot are produced by constructive interference of a monochromatic beam of X-rays scattered
at specific angles from each set of lattice planes in a sample. The atomic positions within the
lattice planes determines intensities of the peak. From the comparison between the pure
commercial LFP powders from MSE supplies and the reclaimed EoL LFP, it is clear that apart
from LFP peaks, there are other impurity peaks (FePO4 and P2O5) observed in the EoL LFP
which can be the decomposition product of LFP after numerous cycles. However, such impurity
peaks disappear in all of the regenerated LFP samples as the intensity of the LFP peaks
apparently grow stronger indicating that aforementioned impurities react with LiOH to form
LFP during the sintering process.
Figure 5-2 The XRD patterns of as-purchased LFP from MSE Supplies, EOL cathode materials,
and recycled cathode materials sintered under 550℃ , 650℃, 750℃ within 𝐀𝐫&𝐇𝟐 atmosphere,
and recycled material sintered under 750℃ in Ar atmosphere.
103
Figure 5-3 The SEM images of (a) as-purchased LFP from MSE Supplies, (b) EOL LFP,
regenerated LFP sintered under (c) 550℃ , (d) 650℃, (e)750℃ within 𝐀𝐫&𝐇𝟐 atmosphere, and
regenerated LFP sintered under (f)750℃ in Ar atmosphere.
The morphology of all six samples are scanned by the environmental SEM (FEI Quanta
600 FEG) provided by Nanoscale Characterization and Fabrication Lab (NCFL) in Virginia
Tech. The FEI Quanta 600 FEG is able to operate under high-vacuum, low-vacuum, and ESEM
modes. The Quanta SEM system is also equipped with analytical systems, energy dispersive
spectrometer, and electron backscatter diffraction. From SEM images in Figure 5-3, large
secondary particles can be observed from the EoL LFP, which is most likely cause by the
existence of the residual PVDF binder. However, these secondary particles does not exist over
regenerated LFP samples regardless of the temperature and atmosphere and the particle
104
distribution in regenerated LFP samples are also more uniformed. Meanwhile, the inert or
reducing atmosphere involved in the sintering process may restrain the thermal decomposition
of the residual PVDF binder and a small amount of PVDF binder might still exist in regenerated
LFP samples. Previous studies 207,208 have pointed out that the residual PVDF have very limited
effects on impedance and the electrochemical performances for the re-assembled cells. Hence,
the existence of PVDF is acceptable for the following coin cell electrochemical performance
tests.
5.2 Coin Cell Assembly
The electrochemical performance of the regenerated LFP cathode material is then tested
in the form of coin cells. LFP cathode electrodes with 14mm in diameter required by coin cells
are produced following the process flow as shown in Figure 5-4. LFP cathode powder and
Timical carbon are first manually dry mixed for 5min. Pre-prepared 5% PVDF/NMP solution
is then added to perform as the binder and the slurry mixture is further mixed for additional
20min. Noticing that within the first two mixing steps, the weight ratio of LFP cathode powder,
carbon, and PVDF need to be controlled at 8:1:1, which is the most commonly adopted ratio in
related research works. A film applicator with a 200μm gap is utilized to coat the slurry onto
the Al current collector foil. After drying in the convection oven for 3 hours under 60℃, the
coated electrode is cut into smaller round electrodes by the precision disc cutter. The last step
before coin cell assembly is to dry the coin cell electrodes in the vacuum oven for additional
12h under 65℃.
105
Figure 5-4 Cathode electrodes preparation for coin cell assembly.
With cathode electrodes available, coin cells are then manually assembled inside the
glove box. The explosive view of the coin cell assembly involved in this study is as shown in
Figure 5-5 (a). The graphite anode electrode usually found in a regular LIB cell is replaced by
a lithium chip for assuring the consistency between cells. Electrolyte, which is 1mol/L LiPF6
dissolves in EC/EMC(3:7 in volume), is dropped between each layer during the assembly
process. The closed positive case and negative case are then press sealed by the MSK-160E
digital coin cell crimper as shown in Figure 5-5 (b). The crimped CR2032 cells as shown in
Figure 5-5 (c) are then tested on the CT2001A classic battery tester.
106
Figure 5-5 (a) Coin cell assembly explosive view, (b) MSK-160E digital coin cell assembly
crimper, (c) assembled CR2032, and (d) CT2001A classic battery tester.
5.3 Electrochemical Performance of the Regenerated LFP
The electrochemical performance of five samples are tested. Apart from four EoL LFP
samples regenerated under different conditions, the as purchased LFP powder from MSE
supplies is also tested as the control experiment to verify the correction of the electrode
preparation process and the coin cell assembly process. The rate performance of the assembled
coin cells are tested at the C rate of 0.1C, 0.2C, 0.5C, and 2C for 5 cycles each. The discharging
cut off voltage is set to 2.6V while the charging cut off voltage is set to 4.0V for all coin cells
107
tested. To avoid the inconsistency caused by the manually conducted electrode preparation
process and the coin cell assembly, three coin cells of each sample are tested and only the ones
with intermediate capacity are selected for comparison.
Figure 5-6 (a) compares the second cycle under 0.1C while Figure 5-6 (b) compares the
result of the rate performance test. It can be seen that 750-Ar&H2 has higher capacity than other
sintered LFP samples at 0.1C. The reversible capacity for the first cycle of 750-Ar&H2 is 129.3
mAh/g while 650-Ar&H2 is 128.6 mAh/g. But the capacity of 650-Ar&H2 start to exceed 750-
Ar&H2 when discharged at 0.2C. At 0.5C and 1C, the capacity of 650-Ar&H2 is higher than
750-Ar&H2 in each cycle. Thus among all sintered materials, 650-Ar&H2 has the best rate
performance as shown in Figure 5-6 (b). Noticing 750-Ar shows the lower specific capacity
and poorer cycle performance compared to 750-Ar&H2 , which indicates that the reducing
atmosphere is required for the LFP sintering process.
Figure 5-6 The charge/discharge voltage profiles at 0.1C (a) and rate performance (b) of cathode
active material mixtures sintered under different conditions.
108
5.4 Conclusion and Future Development
Utilizing the EoL LFP reclaimed from studies of Chapter 4, the cathode materials have
been successfully regenerated by direct solid phase sintering method. The XRD pattern and
SEM images indicate that regenerated samples have a purer phase and more uniformed
morphology compared to reclaimed EoL LFP powders. The capacities of samples regenerated
at 650℃ and 750℃ under Ar&H2 both reach around 130mAh/g at 0.1C discharge conditions.
However, as C rate increases, the performance of 650-Ar&H2 exceeds 750-Ar&H2 , which
might be caused by decomposition of the resynthesized LFP at high regeneration temperature.
The capacity of 750- Ar&H2 exceed 750- Ar over all C-rates indicating that the reducing
atmosphere is necessary for LFP sintering process. Future development of this study may focus
on testing the cycle performance of regenerated LFP samples as well as exploring the influence
of mixing additional LiOH to EoL LFP at various levels under the optimal temperature and
atmosphere.
109
Chapter 6 Summary and Future Work
6.1 Summary of Contributions
The major goal of this dissertation is to facilitate the industrial application of the
pretreatment techniques that is suitable for the direct recycling strategy. Machines and
experiments involving the semi-destructive disassembly approach for cathode electrodes
extraction from the EoL pouch LIBs and the organic solvent approach for the cathode coating
reclamation from the cathode electrodes are designed, conducted, and prototyped.
The first chapter of this dissertation focuses on the complete disassembly strategy of the
EoL pouch LIBs starting from the disassembly sequence planning based on the structure of the
pouch LIB assembly. After successfully allocating all connections in the disassembly
precedence graph, the disassembly system is divided into three consecutive disassembly
modules, namely pouch trimming module, housing removal module, and electrode sorting
module. Customized transporters, fixtures, and end-effectors are designed and prototyped for
each module. The successfully conducted verification tests utilizing dummy cells indicate that
the prototyped automated disassembly system is fully capable of extracting intact cathode
electrodes out of EoL pouch LIBs.
The second chapter of this dissertation moves forward to the subsequent electrode
coating separation process of EoL LCO LIBs focusing on the ultrasonic assisted organic solvent
approach. Placket-Burman parameter screening experiments are first conducted and identified
the DMAC as the best organic solvent while excluding the soaking time, the soaking
temperature, and the sonicating temperature from the list of essential input variables. The
following Taguchi DOE proved that the sonication time, the sheet size, and the solid-liquid
weight ratio are all essential input variables towards the LCO separation yield. The successfully
110
developed linear regression model provides an essential tool for predicting the production
process yield in the future industrial application of the organic solvent approach.
The third chapter studies the organic solvent approach in separating LFP coating
material without involving any external assistance such as ultrasound or brutal impact. The
phenomenon of volume expansion of the LFP cathode coating sheet enables the 100% pure
coating material reclamation without any Al debris from Al current collector. A non-linear
regression model is successfully developed to correlate the input variable (soaking time and
soaking temperature) with the LFP separation yield, providing a reliable approach to predict
the process variables that are able to minimize the energy consumption and maximize the time
efficiency prior to hands-on production process development.
The last part of this dissertation initiates the direct regeneration study of the EoL LFP
cathode coating material reclaimed from the organic solvent approach developed in Chapter 3.
The XRD pattern indicates a successful repair of decomposed LFP powder after the sintering
process and the SEM images shows a much more uniformed particle size distribution in the
regenerated samples. The result of the rate performance test indicate that the optimum process
parameter of the sintering process is 650℃ under the reducing atmosphere.
6.2 Future Work
As aforementioned in section 2.5, the sorting module in the prototyped disassembly
system is upgraded by the vision-sensor network as the first step for developing the cyber-
physical enhanced automated disassembly system. The machine learning algorithm that
cooperates with the physical model of each customized mechanism will fulfill the closed-loop
smart control system with much better process flexibility and situation awareness ability
compared to the originally developed open-loop control UI in LabVIEW. Such improvement
111
will also enable the automated disassembly system to deal with uncertainties from EoL LIBs
due to years of services with more confidence.
The study of the organic solvent approach introduced in Chapter 3 focuses on the LCO
separation yield as the response while putting aside the metal impurity level at the time. As the
subsequent process of the semi-destructive disassembly approach, the studied organic solvent
approach only need to control the impurity level of Al debris since other metals contained in an
EoL LIB such as copper, iron, stainless steel, and etc. are already pre-separated by the
prototyped automated disassembly system. More complete experiment designs considering
both LCO separation yield and Al debris as responses are required. Meanwhile, the intensity
and frequency of the ultrasound need to be taken into consideration in the future study if
condition allows.
The organic solvent approach for separating LFP coating material from the cathode
electrode is studied with 52mm*40mm electrode pieces at a relatively low solid-liquid ratio. In
order to facilitate the industrial-scale application of this approach, future development needs to
focus on treating the entire 0.8m LFP cathode electrode strip. Customized fixtures are needed
to prevent layer stacking while soaking in NMP. Meanwhile, experiments conducted in Chapter
4 utilize fresh NMP for each electrode piece, which can be both expensive and environmentally
harmful for industrial-scale production. Thus the efficiency degradation of reused NMP needs
to be studied systematically in the future development.
Though Chapter 5 reveals the optimum sintering temperature and atmosphere by the
rate performance test, the cycle performance of regenerated samples need to be further
investigated. Meanwhile, the required reducing atmosphere prevents the decomposition of
residual PVDF and the oxidation of carbon black, which may influence the electrochemical
112
performance characterization accuracy of regenerated LFP cathode materials. Thus extra
material separation steps aiming at minimize the residual PVDF and carbon black in the
regenerated LFP cathode material are expected for future development of the direct recycling
research works.
Apart from the aforementioned future works corresponding to research activities of each
Chapter, future studies in adopting Industrial Internet of Things (IIOT) in LIB remanufacturing
process can be a crucial assurance for decreasing variation in the final product as well as
reducing the production cost. Previous study 209 indicated that by applying Advanced Process
Control and streamline the production/material flow, we expect reduced cost of the battery
management system, increased uniformity and yield of individual batteries, and at least 20%
reduction of the manufacturing cost in battery manufacturing processes. Such approach could
benefit the LIB re-manufacturing process even more since much more variations in supply
chains and states of raw material (EoL LIBs) are expected. Thus how IIOT could enhance the
efficiency and effectiveness of the LIB remanufacturing process will be studied in the future
development of this dissertation.
113
REFERENCES
1 Whittingham, M. S. Electrical energy storage and intercalation chemistry. Science 192, 1126-1127
(1976).
2 Mizushima, K., Jones, P., Wiseman, P. & Goodenough, J. B. LixCoO2 (0< x<-1): A new cathode
material for batteries of high energy density. Materials Research Bulletin 15, 783-789 (1980).
3 Yoshino, A., Sanechika, K. & Nakajima, T. (Google Patents, 1987).
4 Prize, T. N. The Nobel Prize in Chemistry 2019. (2019).
5 Fergus, J. W. Recent developments in cathode materials for lithium ion batteries. Journal of power
sources 195, 939-954 (2010).
6 Whittingham, M. S. Lithium batteries and cathode materials. Chemical reviews 104, 4271-4302
(2004).
7 Yazami, R. & Touzain, P. A reversible graphite-lithium negative electrode for electrochemical
generators. Journal of Power Sources 9, 365-371 (1983).
8 Eftekhari, A. Low voltage anode materials for lithium-ion batteries. Energy Storage Materials 7,
157-180 (2017).
9 Hayner, C. M., Zhao, X. & Kung, H. H. Materials for rechargeable lithium-ion batteries. Annual
review of chemical and biomolecular engineering 3, 445-471 (2012).
10 Chawla, N., Bharti, N. & Singh, S. Recent advances in non-flammable electrolytes for safer lithium-
ion batteries. Batteries 5, 19 (2019).
11 Fergus, J. W. Ceramic and polymeric solid electrolytes for lithium-ion batteries. Journal of Power
Sources 195, 4554-4569 (2010).
12 Yao, X. et al. Comparative study of trimethyl phosphite and trimethyl phosphate as electrolyte
additives in lithium ion batteries. Journal of power sources 144, 170-175 (2005).
114
13 Younesi, R., Veith, G. M., Johansson, P., Edström, K. & Vegge, T. Lithium salts for advanced lithium
batteries: Li–metal, Li–O 2, and Li–S. Energy & Environmental Science 8, 1905-1922 (2015).
14 Zhang, S. S. A review on electrolyte additives for lithium-ion batteries. Journal of Power Sources
162, 1379-1394 (2006).
15 Huang, X. & Hitt, J. Lithium ion battery separators: development and performance
characterization of a composite membrane. Journal of membrane science 425, 163-168 (2013).
16 Choi, J.-A., Kim, S. H. & Kim, D.-W. Enhancement of thermal stability and cycling performance in
lithium-ion cells through the use of ceramic-coated separators. Journal of Power Sources 195,
6192-6196 (2010).
17 Abraham, K. M. Exploding Hoverboards Explained. (2016).
18 Guyomard, D. & Tarascon, J. M. Rocking‐chair or lithium‐ion rechargeable lithium batteries.
Advanced Materials 6, 408-412 (1994).
19 Heiskanen, S. K., Kim, J. & Lucht, B. L. Generation and evolution of the solid electrolyte interphase
of lithium-ion batteries. Joule 3, 2322-2333 (2019).
20 Ganesh, P., Kent, P. & Jiang, D.-e. Solid–electrolyte interphase formation and electrolyte reduction
at Li-ion battery graphite anodes: Insights from first-principles molecular dynamics. The Journal
of Physical Chemistry C 116, 24476-24481 (2012).
21 Aurbach, D., Zinigrad, E., Cohen, Y. & Teller, H. A short review of failure mechanisms of lithium
metal and lithiated graphite anodes in liquid electrolyte solutions. Solid state ionics 148, 405-416
(2002).
22 Laresgoiti, I., Käbitz, S., Ecker, M. & Sauer, D. U. Modeling mechanical degradation in lithium ion
batteries during cycling: Solid electrolyte interphase fracture. Journal of Power Sources 300, 112-
122 (2015).
115
23 Lee, S. S., Kim, T. H., Hu, S. J., Cai, W. W. & Abell, J. A. in ASME 2010 international manufacturing
science and engineering conference. 541-549 (American Society of Mechanical Engineers Digital
Collection).
24 Buck, D. S., Fattig, R. N. & Silk, B. J. (Google Patents, 2009).
25 Rao, E. et al. Roll-to-roll functionalization of polyolefin separators for high-performance lithium-
ion batteries. ACS Applied Energy Materials 1, 3292-3300 (2018).
26 Wood III, D. L. et al. Perspectives on the relationship between materials chemistry and roll-to-roll
electrode manufacturing for high-energy lithium-ion batteries. Energy Storage Materials (2020).
27 Kwade, A. et al. Current status and challenges for automotive battery production technologies.
Nature Energy 3, 290-300, doi:10.1038/s41560-018-0130-3 (2018).
28 Mohanty, D. et al. Non-destructive evaluation of slot-die-coated lithium secondary battery
electrodes by in-line laser caliper and IR thermography methods. Analytical Methods 6, 674-683
(2014).
29 Mohanty, D. et al. Effect of electrode manufacturing defects on electrochemical performance of
lithium-ion batteries: Cognizance of the battery failure sources. Journal of Power Sources 312, 70-
79 (2016).
30 Kurfer, J., Westermeier, M., Tammer, C. & Reinhart, G. Production of large-area lithium-ion cells–
Preconditioning, cell stacking and quality assurance. CIRP annals 61, 1-4 (2012).
31 Zhang, S. S. A review on the separators of liquid electrolyte Li-ion batteries. Journal of power
sources 164, 351-364 (2007).
32 Schilling, A., Schmitt, J., Dietrich, F. & Dröder, K. Analyzing Bending Stresses on Lithium‐Ion Battery
Cathodes induced by the Assembly Process. Energy Technology 4, 1502-1508 (2016).
116
33 Schmitt, J., Raatz, A., Dietrich, F., Dröder, K. & Hesselbach, J. Process and performance
optimization by selective assembly of battery electrodes. CIRP Annals 63, 9-12,
doi:10.1016/j.cirp.2014.03.018 (2014).
34 Knoche, T., Surek, F. & Reinhart, G. A process model for the electrolyte filling of lithium-ion
batteries. Procedia CIRP 41, 405-410 (2016).
35 Lu, P., Li, C., Schneider, E. W. & Harris, S. J. Chemistry, impedance, and morphology evolution in
solid electrolyte interphase films during formation in lithium ion batteries. The Journal of Physical
Chemistry C 118, 896-903 (2014).
36 Bridges, C. A., Sun, X.-G., Zhao, J., Paranthaman, M. P. & Dai, S. In situ observation of solid
electrolyte interphase formation in ordered mesoporous hard carbon by small-angle neutron
scattering. The Journal of Physical Chemistry C 116, 7701-7711 (2012).
37 Imhof, R. & Novák, P. In situ investigation of the electrochemical reduction of carbonate
electrolyte solutions at graphite electrodes. Journal of The Electrochemical Society 145, 1081-
1087 (1998).
38 Schröder, R., Aydemir, M. & Seliger, G. Comparatively Assessing different Shapes of Lithium-ion
Battery Cells. Procedia Manufacturing 8, 104-111, doi:10.1016/j.promfg.2017.02.013 (2017).
39 Kwade, A. et al. Current status and challenges for automotive battery production technologies.
Nature Energy 3, 290 (2018).
40 Lu, L., Han, X., Li, J., Hua, J. & Ouyang, M. A review on the key issues for lithium-ion battery
management in electric vehicles. Journal of power sources 226, 272-288 (2013).
41 Goodenough, J. B. & Kim, Y. Challenges for rechargeable Li batteries. Chemistry of materials 22,
587-603 (2010).
42 Vetter, J. et al. Ageing mechanisms in lithium-ion batteries. Journal of power sources 147, 269-
281 (2005).
117
43 Jannesari, H., Emami, M. & Ziegler, C. Effect of electrolyte transport properties and variations in
the morphological parameters on the variation of side reaction rate across the anode electrode
and the aging of lithium ion batteries. Journal of Power Sources 196, 9654-9664 (2011).
44 Song, M. & Choe, S.-Y. Fast and safe charging method suppressing side reaction and lithium
deposition reaction in lithium ion battery. Journal of Power Sources 436, 226835 (2019).
45 Arora, P., White, R. E. & Doyle, M. Capacity fade mechanisms and side reactions in lithium‐ion
batteries. Journal of the Electrochemical Society 145, 3647 (1998).
46 Barré, A. et al. A review on lithium-ion battery ageing mechanisms and estimations for automotive
applications. Journal of Power Sources 241, 680-689 (2013).
47 Wohlfahrt-Mehrens, M., Vogler, C. & Garche, J. Aging mechanisms of lithium cathode materials.
Journal of power sources 127, 58-64 (2004).
48 Deshpande, R. D. & Bernardi, D. M. Modeling solid-electrolyte interphase (SEI) fracture: coupled
mechanical/chemical degradation of the lithium ion battery. Journal of The Electrochemical
Society 164, A461 (2017).
49 Shin, H., Park, J., Sastry, A. M. & Lu, W. Degradation of the solid electrolyte interphase induced by
the deposition of manganese ions. Journal of Power Sources 284, 416-427 (2015).
50 Kim, T., Ono, L. K., Fleck, N., Raga, S. R. & Qi, Y. Transition metal speciation as a degradation
mechanism with the formation of a solid-electrolyte interphase (SEI) in Ni-rich transition metal
oxide cathodes. Journal of Materials Chemistry A 6, 14449-14463 (2018).
51 Tang, C.-Y., Ma, Y., Haasch, R. T., Ouyang, J.-H. & Dillon, S. J. Insights into solid-electrolyte
interphase induced Li-ion degradation from in situ auger electron spectroscopy. The journal of
physical chemistry letters 8, 6226-6230 (2017).
118
52 Kim, J., Kim, H., Ryu, J. H. & Oh, S. M. Communication—Lithium Bis (fluorosulfonyl) imide (LiFSI)
as a Promising Salt to Suppress Solid Electrolyte Interphase Degradation at Elevated Temperatures.
Journal of The Electrochemical Society 167, 080529 (2020).
53 Balbuena, P. B. & Wang, Y. Lithium-ion batteries: solid-electrolyte interphase. (Imperial college
press, 2004).
54 Liu, G. & Lu, W. A model of concurrent lithium dendrite growth, SEI growth, SEI penetration and
regrowth. Journal of the Electrochemical Society 164, A1826 (2017).
55 Deshpande, R., Verbrugge, M., Cheng, Y.-T., Wang, J. & Liu, P. Battery cycle life prediction with
coupled chemical degradation and fatigue mechanics. Journal of the Electrochemical Society 159,
A1730 (2012).
56 Li, Z., Huang, J., Liaw, B. Y., Metzler, V. & Zhang, J. A review of lithium deposition in lithium-ion
and lithium metal secondary batteries. Journal of power sources 254, 168-182 (2014).
57 Perkins, R. D., Randall, A. V., Zhang, X. & Plett, G. L. Controls oriented reduced order modeling of
lithium deposition on overcharge. Journal of Power Sources 209, 318-325 (2012).
58 Aurbach, D. Review of selected electrode–solution interactions which determine the performance
of Li and Li ion batteries. Journal of Power Sources 89, 206-218 (2000).
59 Koltypin, M., Cohen, Y. S., Markovsky, B., Cohen, Y. & Aurbach, D. The study of lithium insertion–
deinsertion processes into composite graphite electrodes by in situ atomic force microscopy
(AFM). Electrochemistry communications 4, 17-23 (2002).
60 Yamaki, J.-i. et al. A consideration of the morphology of electrochemically deposited lithium in an
organic electrolyte. Journal of Power Sources 74, 219-227 (1998).
61 Wen, J., Yu, Y. & Chen, C. A review on lithium-ion batteries safety issues: existing problems and
possible solutions. Materials express 2, 197-212 (2012).
119
62 Chen, Z., Christensen, L. & Dahn, J. Comparison of PVDF and PVDF-TFE-P as binders for electrode
materials showing large volume changes in lithium-ion batteries. Journal of The Electrochemical
Society 150, A1073 (2003).
63 Voelker, P. Trace degradation analysis of lithium-ion battery components. R&D Magazinde,(April
2014) http://www. rdmag. com/articles/2014/04/trace-degradation-analysislithium-ion-battery-
components Search PubMed,(accessed August 2015) (2014).
64 Abe, K. et al. Additives-containing functional electrolytes for suppressing electrolyte
decomposition in lithium-ion batteries. Electrochimica acta 49, 4613-4622 (2004).
65 Andersson, A. M., Edström, K. & Thomas, J. O. Characterisation of the ambient and elevated
temperature performance of a graphite electrode. Journal of power sources 81, 8-12 (1999).
66 Ramadass, P., Haran, B., White, R. & Popov, B. N. Capacity fade of Sony 18650 cells cycled at
elevated temperatures: Part I. Cycling performance. Journal of power sources 112, 606-613 (2002).
67 Abraham, D. et al. Surface changes on LiNi0. 8Co0. 2O2 particles during testing of high-power
lithium-ion cells. Electrochemistry communications 4, 620-625 (2002).
68 Venkatraman, S., Shin, Y. & Manthiram, A. Phase Relationships and Structural and Chemical
Stabilities of Charged Li1− x CoO2− δ and Li1− x Ni0. 85Co0. 15 O 2− δ Cathodes. Electrochemical
and solid-state letters 6, A9-A12 (2003).
69 Park, J., Seo, J. H., Plett, G., Lu, W. & Sastry, A. M. Numerical simulation of the effect of the
dissolution of LiMn2O4 particles on Li-ion battery performance. Electrochemical and Solid-State
Letters 14, A14-A18 (2011).
70 Bandhauer, T. M., Garimella, S. & Fuller, T. F. A critical review of thermal issues in lithium-ion
batteries. Journal of the Electrochemical Society 158, R1-R25 (2011).
71 Wang, X., Gaustad, G., Babbitt, C. W. & Richa, K. Economies of scale for future lithium-ion battery
recycling infrastructure. Resources, Conservation and Recycling 83, 53-62 (2014).
120
72 Bernardes, A. M., Espinosa, D. C. R. & Tenório, J. S. Recycling of batteries: a review of current
processes and technologies. Journal of Power Sources 130, 291-298 (2004).
73 Li, Y. et al. Leaching of heavy metals from E-waste in simulated landfill columns. Waste
Management 29, 2147-2150 (2009).
74 Yun, L. et al. Metallurgical and mechanical methods for recycling of lithium-ion battery pack for
electric vehicles. Resources, Conservation and Recycling 136, 198-208 (2018).
75 Winslow, K. M., Laux, S. J. & Townsend, T. G. A review on the growing concern and potential
management strategies of waste lithium-ion batteries. Resources, Conservation and Recycling 129,
263-277 (2018).
76 Meshram, P., Pandey, B. & Mankhand, T. Recovery of valuable metals from cathodic active
material of spent lithium ion batteries: Leaching and kinetic aspects. Waste management 45, 306-
313 (2015).
77 Xu, J. et al. A review of processes and technologies for the recycling of lithium-ion secondary
batteries. Journal of Power Sources 177, 512-527, doi:10.1016/j.jpowsour.2007.11.074 (2008).
78 Hu, J., Zhang, J., Li, H., Chen, Y. & Wang, C. A promising approach for the recovery of high value-
added metals from spent lithium-ion batteries. Journal of Power Sources 351, 192-199 (2017).
79 Gratz, E., Sa, Q., Apelian, D. & Wang, Y. A closed loop process for recycling spent lithium ion
batteries. Journal of Power Sources 262, 255-262 (2014).
80 Yang, T. et al. An Effective Relithiation Process for Recycling Lithium‐Ion Battery Cathode
Materials. Advanced Sustainable Systems 4, 1900088 (2020).
81 Li, X., Zhang, J., Song, D., Song, J. & Zhang, L. Direct regeneration of recycled cathode material
mixture from scrapped LiFePO4 batteries. Journal of Power Sources 345, 78-84 (2017).
82 Lv, W. et al. A Critical Review and Analysis on the Recycling of Spent Lithium-Ion Batteries. ACS
Sustainable Chemistry & Engineering 6, 1504-1521, doi:10.1021/acssuschemeng.7b03811 (2018).
121
83 Harper, G. et al. Recycling lithium-ion batteries from electric vehicles. Nature 575, 75-86 (2019).
84 Pistoia, G. & Liaw, B. Behaviour of Lithium-Ion Batteries in Electric Vehicles: Battery Health,
Performance, Safety, and Cost. (Springer, 2018).
85 Herrmann, C., Raatz, A., Andrew, S. & Schmitt, J. Scenario-Based Development of Disassembly
Systems for Automotive Lithium Ion Battery Systems. Advanced Materials Research 907, 391-401,
doi:10.4028/www.scientific.net/AMR.907.391 (2014).
86 Adhikaree, A. et al. in 2017 IEEE Energy Conversion Congress and Exposition (ECCE). 1004-1009
(IEEE).
87 Maharshi, S. Cloud based disassembly of electric vehicle battery. Procedia Manufacturing 30, 136-
142 (2019).
88 Schmitt, J., Haupt, H., Kurrat, M. & Raatz, A. in 2011 15th International Conference on Advanced
Robotics (ICAR). 291-297 (IEEE).
89 Wegener, K., Andrew, S., Raatz, A., Dröder, K. & Herrmann, C. Disassembly of electric vehicle
batteries using the example of the Audi Q5 hybrid system. Procedia CIRP 23, 155-160 (2014).
90 Wegener, K., Chen, W. H., Dietrich, F., Dröder, K. & Kara, S. Robot assisted disassembly for the
recycling of electric vehicle batteries. Procedia Cirp 29, 716-721 (2015).
91 Luo, J., Solowjow, E., Wen, C., Ojea, J. A. & Agogino, A. M. in 2018 IEEE/RSJ International
Conference on Intelligent Robots and Systems (IROS). 2062-2069 (IEEE).
92 Inoue, T., De Magistris, G., Munawar, A., Yokoya, T. & Tachibana, R. in 2017 IEEE/RSJ International
Conference on Intelligent Robots and Systems (IROS). 819-825 (IEEE).
93 Nerakae, P., Uangpairoj, P. & Chamniprasart, K. Using machine vision for flexible automatic
assembly system. Procedia Computer Science 96, 428-435 (2016).
94 Öhl, J., Horn, D., Zimmermann, J., Stauber, R. & Gutfleisch, O. in Materials Science Forum. 74-78
(Trans Tech Publ).
122
95 Shin, S. M., Kim, N. H., Sohn, J. S., Yang, D. H. & Kim, Y. H. Development of a metal recovery process
from Li-ion battery wastes. Hydrometallurgy 79, 172-181 (2005).
96 Pagnanelli, F., Moscardini, E., Altimari, P., Atia, T. A. & Toro, L. Cobalt products from real waste
fractions of end of life lithium ion batteries. Waste management 51, 214-221 (2016).
97 Hanisch, C. et al. Recycling of lithium-ion batteries: a novel method to separate coating and foil of
electrodes. Journal of cleaner production 108, 301-311 (2015).
98 Zeng, X., Li, J. & Shen, B. Novel approach to recover cobalt and lithium from spent lithium-ion
battery using oxalic acid. Journal of hazardous materials 295, 112-118 (2015).
99 Zhang, X. et al. Toward sustainable and systematic recycling of spent rechargeable batteries.
Chem Soc Rev 47, 7239-7302, doi:10.1039/c8cs00297e (2018).
100 Hess, S., Wohlfahrt-Mehrens, M. & Wachtler, M. Flammability of Li-ion battery electrolytes: flash
point and self-extinguishing time measurements. Journal of The Electrochemical Society 162,
A3084 (2015).
101 Wang, Q., Sun, J. & Chu, G. Lithium ion battery fire and explosion. Fire Safety Science 8, 375-382
(2005).
102 Diekmann, J. et al. Ecological recycling of lithium-ion batteries from electric vehicles with focus on
mechanical processes. Journal of the Electrochemical Society 164, A6184-A6191 (2017).
103 Diekmann, J., Rothermel, S., Nowak, S. & Kwade, A. in Recycling of Lithium-Ion Batteries 33-38
(Springer, 2018).
104 Diekmann, J., Hanisch, C., Loellhoeffel, T., Schälicke, G. & Kwade, A. Ecologically Friendly Recycling
of Lithium-Ion Batteries-the LithoRec Process. ECS Transactions 73, 1 (2016).
105 Yan, J. Handbook of Clean Energy Systems, 6 Volume Set. Vol. 5 (John Wiley & Sons, 2015).
123
106 Diekmann, J. et al. Ecological Recycling of Lithium-Ion Batteries from Electric Vehicles with Focus
on Mechanical Processes. Journal of The Electrochemical Society 164, A6184-A6191,
doi:10.1149/2.0271701jes (2016).
107 Contestabile, M., Panero, S. & Scrosati, B. A laboratory-scale lithium-ion battery recycling process.
Journal of Power Sources 92, 65-69 (2001).
108 Hanisch, C., Haselrieder, W. & Kwade, A. in Glocalized Solutions for Sustainability in Manufacturing
85-89 (Springer, 2011).
109 Hanisch, C. et al. In-production recycling of active materials from lithium-ion battery scraps. ECS
Transactions 64, 131 (2015).
110 Hanisch, C., Haselrieder, W. & Kwade, A. (Google Patents, 2017).
111 Li, J., Shi, P., Wang, Z., Chen, Y. & Chang, C.-C. A combined recovery process of metals in spent
lithium-ion batteries. Chemosphere 77, 1132-1136 (2009).
112 Li, L., Chen, R., Sun, F., Wu, F. & Liu, J. Preparation of LiCoO2 films from spent lithium-ion batteries
by a combined recycling process. Hydrometallurgy 108, 220-225 (2011).
113 Li, L. et al. Ascorbic-acid-assisted recovery of cobalt and lithium from spent Li-ion batteries.
Journal of Power Sources 218, 21-27 (2012).
114 Song, D. et al. Heat treatment of LiCoO2 recovered from cathode scraps with solvent method.
Journal of Power Sources 249, 137-141 (2014).
115 Song, X. et al. Direct regeneration of cathode materials from spent lithium iron phosphate
batteries using a solid phase sintering method. RSC advances 7, 4783-4790 (2017).
116 Toma, C. et al. in IOP Conference Series: Materials Science and Engineering. 012034 (IOP
Publishing).
117 Fouad, O., Farghaly, F. & Bahgat, M. A novel approach for synthesis of nanocrystalline γ-LiAlO2
from spent lithium-ion batteries. Journal of analytical and applied pyrolysis 78, 65-69 (2007).
124
118 Bahgat, M., Farghaly, F., Basir, S. A. & Fouad, O. Synthesis, characterization and magnetic
properties of microcrystalline lithium cobalt ferrite from spent lithium-ion batteries. Journal of
materials processing technology 183, 117-121 (2007).
119 Sun, Z. et al. Toward sustainability for recovery of critical metals from electronic waste: the
hydrochemistry processes. ACS Sustainable Chemistry & Engineering 5, 21-40 (2017).
120 Xiao, J., Li, J. & Xu, Z. Recycling metals from lithium ion battery by mechanical separation and
vacuum metallurgy. Journal of hazardous materials 338, 124-131 (2017).
121 Georgi-Maschler, T., Friedrich, B., Weyhe, R., Heegn, H. & Rutz, M. Development of a recycling
process for Li-ion batteries. Journal of power sources 207, 173-182 (2012).
122 Yao, Y. et al. Hydrometallurgical processes for recycling spent lithium-ion batteries: a critical
review. ACS Sustainable Chemistry & Engineering 6, 13611-13627 (2018).
123 Larcher, D. & Tarascon, J.-M. Towards greener and more sustainable batteries for electrical energy
storage. Nature chemistry 7, 19-29 (2015).
124 Chagnes, A. & Pospiech, B. A brief review on hydrometallurgical technologies for recycling spent
lithium‐ion batteries. Journal of Chemical Technology & Biotechnology 88, 1191-1199 (2013).
125 Li, L. et al. Sustainable recovery of cathode materials from spent lithium-ion batteries using lactic
acid leaching system. ACS Sustainable Chemistry & Engineering 5, 5224-5233 (2017).
126 Nayaka, G. P. et al. An environmental friendly attempt to recycle the spent Li-ion battery cathode
through organic acid leaching. Journal of Environmental Chemical Engineering 7,
doi:10.1016/j.jece.2018.102854 (2019).
127 Song, D. et al. Recovery and heat treatment of the Li (Ni1/3Co1/3Mn1/3) O2 cathode scrap
material for lithium ion battery. Journal of power sources 232, 348-352 (2013).
125
128 Mantuano, D. P., Dorella, G., Elias, R. C. A. & Mansur, M. B. Analysis of a hydrometallurgical route
to recover base metals from spent rechargeable batteries by liquid–liquid extraction with Cyanex
272. Journal of Power Sources 159, 1510-1518 (2006).
129 Tanong, K., Coudert, L., Mercier, G. & Blais, J.-F. Recovery of metals from a mixture of various
spent batteries by a hydrometallurgical process. Journal of environmental management 181, 95-
107 (2016).
130 Wang, F., Sun, R., Xu, J., Chen, Z. & Kang, M. Recovery of cobalt from spent lithium ion batteries
using sulphuric acid leaching followed by solid–liquid separation and solvent extraction. RSC
advances 6, 85303-85311 (2016).
131 Wang, R.-C., Lin, Y.-C. & Wu, S.-H. A novel recovery process of metal values from the cathode
active materials of the lithium-ion secondary batteries. Hydrometallurgy 99, 194-201 (2009).
132 Barik, S., Prabaharan, G. & Kumar, L. Leaching and separation of Co and Mn from electrode
materials of spent lithium-ion batteries using hydrochloric acid: Laboratory and pilot scale study.
Journal of Cleaner Production 147, 37-43 (2017).
133 Takacova, Z., Havlik, T., Kukurugya, F. & Orac, D. Cobalt and lithium recovery from active mass of
spent Li-ion batteries: Theoretical and experimental approach. Hydrometallurgy 163, 9-17 (2016).
134 Barbieri, E., Lima, E., Cantarino, S., Lelis, M. & Freitas, M. Recycling of spent ion-lithium batteries
as cobalt hydroxide, and cobalt oxide films formed under a conductive glass substrate, and their
electrochemical properties. Journal of Power Sources 269, 158-163 (2014).
135 Barbieri, E., Lima, E., Lelis, M. & Freitas, M. Recycling of cobalt from spent Li-ion batteries as β-Co
(OH) 2 and the application of Co3O4 as a pseudocapacitor. Journal of Power Sources 270, 158-165
(2014).
136 Lee, C. K. & Rhee, K.-I. Reductive leaching of cathodic active materials from lithium ion battery
wastes. Hydrometallurgy 68, 5-10 (2003).
126
137 Castillo, S., Ansart, F., Laberty-Robert, C. & Portal, J. Advances in the recovering of spent lithium
battery compounds. Journal of Power Sources 112, 247-254 (2002).
138 Li, J. et al. Study of extraction and purification of Ni, Co and Mn from spent battery material.
Hydrometallurgy 99, 7-12 (2009).
139 Meshram, P., Pandey, B. D., Mankhand, T. R. & Deveci, H. Comparision of different reductants in
leaching of spent lithium ion batteries. Jom 68, 2613-2623 (2016).
140 Pagnanelli, F. et al. Acid reducing leaching of cathodic powder from spent lithium ion batteries:
glucose oxidative pathways and particle area evolution. Journal of Industrial and Engineering
Chemistry 20, 3201-3207 (2014).
141 Yang, L. et al. Preparation and magnetic performance of Co0. 8Fe2. 2O4 by a sol–gel method using
cathode materials of spent Li-ion batteries. Ceramics International 42, 1897-1902 (2016).
142 Nayaka, G., Pai, K., Santhosh, G. & Manjanna, J. Dissolution of cathode active material of spent Li-
ion batteries using tartaric acid and ascorbic acid mixture to recover Co. Hydrometallurgy 161, 54-
57 (2016).
143 Pinna, E. G., Ruiz, M. C., Ojeda, M. W. & Rodriguez, M. H. Cathodes of spent Li-ion batteries:
Dissolution with phosphoric acid and recovery of lithium and cobalt from leach liquors.
Hydrometallurgy 167, 66-71 (2017).
144 Ferreira, D. A., Prados, L. M. Z., Majuste, D. & Mansur, M. B. Hydrometallurgical separation of
aluminium, cobalt, copper and lithium from spent Li-ion batteries. Journal of Power Sources 187,
238-246 (2009).
145 Zheng, Y. et al. Leaching procedure and kinetic studies of cobalt in cathode materials from spent
lithium ion batteries using organic citric acid as leachant. International Journal of Environmental
Research 10, 159-168 (2016).
127
146 Sun, L. & Qiu, K. Organic oxalate as leachant and precipitant for the recovery of valuable metals
from spent lithium-ion batteries. Waste management 32, 1575-1582 (2012).
147 Liang, Y. et al. Life cycle assessment of lithium-ion batteries for greenhouse gas emissions.
Resources, conservation and recycling 117, 285-293 (2017).
148 Yang, L., Xi, G. & Xi, Y. Recovery of Co, Mn, Ni, and Li from spent lithium ion batteries for the
preparation of LiNixCoyMnzO2 cathode materials. Ceramics international 41, 11498-11503 (2015).
149 Cai, G., Fung, K. Y., Ng, K. M. & Wibowo, C. Process development for the recycle of spent lithium
ion batteries by chemical precipitation. Industrial & Engineering Chemistry Research 53, 18245-
18259 (2014).
150 Joulié, M., Laucournet, R. & Billy, E. Hydrometallurgical process for the recovery of high value
metals from spent lithium nickel cobalt aluminum oxide based lithium-ion batteries. Journal of
Power Sources 247, 551-555 (2014).
151 He, L.-P., Sun, S.-Y., Song, X.-F. & Yu, J.-G. Leaching process for recovering valuable metals from
the LiNi1/3Co1/3Mn1/3O2 cathode of lithium-ion batteries. Waste management 64, 171-181
(2017).
152 Li, L. et al. Recovery of valuable metals from spent lithium-ion batteries by ultrasonic-assisted
leaching process. Journal of Power Sources 262, 380-385 (2014).
153 Nie, H. et al. LiCoO 2: recycling from spent batteries and regeneration with solid state synthesis.
Green chemistry 17, 1276-1280 (2015).
154 Chen, J. et al. Environmentally friendly recycling and effective repairing of cathode powders from
spent LiFePO 4 batteries. Green Chemistry 18, 2500-2506 (2016).
155 Chen, S. et al. Renovation of LiCoO2 with outstanding cycling stability by thermal treatment with
Li2CO3 from spent Li-ion batteries. Journal of Energy Storage 8, 262-273 (2016).
128
156 Shi, Y., Chen, G. & Chen, Z. Effective regeneration of LiCoO2 from spent lithium-ion batteries: a
direct approach towards high-performance active particles. Green Chemistry 20, 851-862,
doi:10.1039/c7gc02831h (2018).
157 Dewulf, J. et al. Recycling rechargeable lithium ion batteries: Critical analysis of natural resource
savings. Resources, Conservation and Recycling 54, 229-234 (2010).
158 Zou, H., Gratz, E., Apelian, D. & Wang, Y. A novel method to recycle mixed cathode materials for
lithium ion batteries. Green Chemistry 15, 1183-1191 (2013).
159 Zhang, T. et al. Surface analysis of cobalt-enriched crushed products of spent lithium-ion batteries
by X-ray photoelectron spectroscopy. Separation and Purification Technology 138, 21-27 (2014).
160 Herrmann, C., Raatz, A., Mennenga, M., Schmitt, J. & Andrew, S. in Leveraging Technology for a
Sustainable World 149-154 (Springer, 2012).
161 Weyrich, M., Natkunarajah, N. & Sc, M. Conception of an automated plant for the disassembly of
lithium ion batteries. Germany, Paul-Bonatz-Str, 9-11.
162 Zhang, T. et al. Chemical and process mineralogical characterizations of spent lithium-ion
batteries: an approach by multi-analytical techniques. Waste management 34, 1051-1058 (2014).
163 Sa, Q. et al. Synthesis of high performance LiNi1/3Mn1/3Co1/3O2 from lithium ion battery
recovery stream. Journal of Power Sources 282, 140-145 (2015).
164 Guo, L. & Liu, G. W. in Advanced Materials Research. 983-988 (Trans Tech Publ).
165 Zhang, G. Q., Li, L. M. & Choi, S. in 2012 IEEE International Conference on Technologies for Practical
Robot Applications (TePRA). 117-120 (IEEE).
166 Yao, L., Xi, Y., Xi, G. & Feng, Y. Synthesis of cobalt ferrite with enhanced magnetostriction
properties by the sol− gel− hydrothermal route using spent Li-ion battery. Journal of Alloys and
Compounds 680, 73-79 (2016).
129
167 Santana, I., Moreira, T., Lelis, M. & Freitas, M. Photocatalytic properties of Co3O4/LiCoO2 recycled
from spent lithium-ion batteries using citric acid as leaching agent. Materials Chemistry and
Physics 190, 38-44 (2017).
168 Yang, Z. et al. Different types of MnO 2 recovered from spent LiMn 2 O 4 batteries and their
application in electrochemical capacitors. Journal of Materials Science 48, 2512-2519 (2013).
169 Moura, M. et al. Synthesis, characterization and photocatalytic properties of nanostructured
CoFe2O4 recycled from spent Li-ion batteries. Chemosphere 182, 339-347 (2017).
170 Senćanski, J. et al. The synthesis of Li (CoMnNi) O2 cathode material from spent-Li ion batteries
and the proof of its functionality in aqueous lithium and sodium electrolytic solutions. Journal of
Power Sources 342, 690-703 (2017).
171 Perez, E. et al. Recovery of metals from simulant spent lithium-ion battery as organophosphonate
coordination polymers in aqueous media. Journal of hazardous materials 317, 617-621 (2016).
172 Curry, C. Lithium-ion battery costs and market. Bloomberg New Energy Finance 5 (2017).
173 Peterson, S. B., Apt, J. & Whitacre, J. Lithium-ion battery cell degradation resulting from realistic
vehicle and vehicle-to-grid utilization. Journal of Power Sources 195, 2385-2392 (2010).
174 Ellingsen, L. A. W. et al. Life cycle assessment of a lithium‐ion battery vehicle pack. Journal of
Industrial Ecology 18, 113-124 (2014).
175 Sun, X., Hao, H., Zhao, F. & Liu, Z. Tracing global lithium flow: A trade-linked material flow analysis.
Resources, Conservation and Recycling 124, 50-61 (2017).
176 Zhang, X. et al. Toward sustainable and systematic recycling of spent rechargeable batteries.
Chemical Society Reviews 47, 7239-7302 (2018).
177 Shi, Y., Chen, G. & Chen, Z. Effective regeneration of LiCoO 2 from spent lithium-ion batteries: a
direct approach towards high-performance active particles. Green Chemistry 20, 851-862 (2018).
130
178 Kim, D.-S. et al. Simultaneous separation and renovation of lithium cobalt oxide from the cathode
of spent lithium ion rechargeable batteries. Journal of Power Sources 132, 145-149 (2004).
179 Zheng, L., Sturges, R., Li, L. & Yang, T.
180 Schröder, R., Glodde, A., Aydemir, M. & Bach, G. in Applied Mechanics and Materials. 19-26 (Trans
Tech Publ).
181 LeCun, Y., Bengio, Y. & Hinton, G. Deep learning. nature 521, 436-444 (2015).
182 Zheng, H., Fu, J., Mei, T. & Luo, J. in Proceedings of the IEEE international conference on computer
vision. 5209-5217.
183 Farrugia, H. Convolutional Neural Networks (CNNs) and Use Cases in Health. (2019).
184 Meshram, P., Pandey, B. & Mankhand, T. Extraction of lithium from primary and secondary
sources by pre-treatment, leaching and separation: A comprehensive review. Hydrometallurgy
150, 192-208 (2014).
185 Tedjar, F. Challenges for recycling advanced Li‐ion batteries. (2013).
186 Katwala, A. The spiralling environmental cost of our lithium battery addiction. . (2018).
187 Els, P. Will lithium-ion battery supply slow the EV revolution? (2019).
188 Eckart, J. Batteries can be part of the fight against climate change - if we do these five things.
(2017).
189 Xu, J. et al. A review of processes and technologies for the recycling of lithium-ion secondary
batteries. Journal of Power Sources 177, 512-527 (2008).
190 Li, L. et al. Disassembly Automation for Recycling End-of-Life Lithium-Ion Pouch Cells. JOM 71,
4457-4464 (2019).
191 Li, L. et al. Disassembly Automation for Recycling End-of-Life Lithium-Ion Pouch Cells. JOM, 1-8
(2019).
192 Kendall, M. G. The advanced theory of statistics. The advanced theory of statistics. (1946).
131
193 Neyman, J. Outline of a theory of statistical estimation based on the classical theory of probability.
Philosophical Transactions of the Royal Society of London. Series A, Mathematical and Physical
Sciences 236, 333-380 (1937).
194 Bland, J. M. & Peacock, J. L. Interpreting statistics with confidence. The Obstetrician &
Gynaecologist 4, 176-180 (2002).
195 Du Prel, J.-B., Hommel, G., Röhrig, B. & Blettner, M. Confidence interval or p-value?: part 4 of a
series on evaluation of scientific publications. Deutsches Ärzteblatt International 106, 335 (2009).
196 Hedayat, A. S., Sloane, N. J. A. & Stufken, J. Orthogonal arrays: theory and applications. (Springer
Science & Business Media, 2012).
197 Taguchi, G. & Konishi, S. Orthogonal arrays and linear graphs. (na, 1987).
198 Sloane, N. J. A library of orthogonal arrays. Fixed-level arrays with more than three levels: OA 16
(2007).
199 Wang, W. & Wu, Y. An overview of recycling and treatment of spent LiFePO4 batteries in China.
Resources, Conservation and Recycling 127, 233-243 (2017).
200 McClenathan, D. M., Wetzel, W. C., Lorge, S. E. & Hieftje, G. M. Effect of the plasma operating
frequency on the figures of merit of an inductively coupled plasma time-of-flight mass
spectrometer. Journal of Analytical Atomic Spectrometry 21, 160-167 (2006).
201 Sloop, S. et al. A direct recycling case study from a lithium-ion battery recall. Sustainable Materials
and Technologies 25, e00152 (2020).
202 Gao, Y., Li, Y., Li, J., Xie, H. & Chen, Y. Direct recovery of LiCoO2 from the recycled lithium-ion
batteries via structure restoration. Journal of Alloys and Compounds, 156234 (2020).
203 Liu, Y.-J., Hu, Q.-Y., Li, X.-H., Wang, Z.-X. & Guo, H.-J. Recycle and synthesis of LiCoO2 from incisors
bound of Li-ion batteries. Transactions of Nonferrous Metals Society of China 16, 956-959 (2006).
132
204 Zheng, P. The Design and Optimization of a Lithium-ion Battery Direct Recycling Process, Virginia
Tech, (2019).
205 Ge, D. Direct Lithium-ion Battery Recycling to Yield Battery Grade Cathode Materials, Virginia Tech,
(2019).
206 Jie, Y. et al. in REWAS 2019 401-409 (Springer, 2019).
207 Song, X. et al. Direct regeneration of cathode materials from spent lithium iron phosphate
batteries using a solid phase sintering method. RSC Advances 7, 4783-4790,
doi:10.1039/c6ra27210j (2017).
208 Natarajan, S., Lakshmi, D. S., Bajaj, H. C. & Srivastava, D. N. Recovery and utilization of graphite
and polymer materials from spent lithium-ion batteries for synthesizing polymer–graphite
nanocomposite thin films. Journal of Environmental Chemical Engineering 3, 2538-2545 (2015).
209 Asif, A. A. & Singh, R. Further cost reduction of battery manufacturing. Batteries 3, 17 (2017).
Top Related